Citation
Characteristics of wildlife highway-crossing zones and applying this information to reduce wildlife/highway conflicts

Material Information

Title:
Characteristics of wildlife highway-crossing zones and applying this information to reduce wildlife/highway conflicts
Creator:
Barnum, Sarah
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
159 leaves : ill. ; 28 cm.

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
College of Architecture and Planning, CU Denver
Degree Disciplines:
Design and planning

Subjects

Subjects / Keywords:
Wildlife crossings ( lcsh )
Traffic safety and wildlife ( lcsh )
Roadkill ( lcsh )
Roadside animals ( lcsh )
Roads -- Environmental aspects ( lcsh )
Roadkill ( fast )
Roads -- Environmental aspects ( fast )
Roadside animals ( fast )
Traffic safety and wildlife ( fast )
Wildlife crossings ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Colorado at Denver, 2003. Design and planning
Bibliography:
Includes bibliographical references (leaves 149-159).
General Note:
College of Architecture and Planning
Statement of Responsibility:
by Sarah Barnum.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
54671166 ( OCLC )
ocm54671166

Downloads

This item has the following downloads:


Full Text
CHARACTERISTICS OF WILDLIFE HIGHWAY-CROSSING ZONES AND
APPLYING THIS INFORMATION TO REDUCE WILDLIFE/HIGHWAY CONFLICTS
by
Sarah Bamum
B.S., University of Vermont, 1988 M.S., Utah State University, 1994
A thesis submitted to the University of Colorado at Denver in partial fulfillment of the requirements for the degree of Doctor of Philosophy Design and Planning


This dissertation for the Doctor of Philosophy Degree by Sarah A. Baraum
has been approved by
Willem Van Vliet -
James Huff
i35. S.OoZ


Bamum, Sarah A. (Ph.D., Design and Planning)
Characteristics of Wildlife Highway-Crossing Zones and Applying This Information to Reduce Wildlife/Highway Conflicts
Dissertation directed by Professor Willem K. T. Van Vliet
ABSTRACT
Negative impacts of highways on wildlife include highway-induced habitat loss and alterations, habitat fragmentation, direct mortality, and disturbance.
Although the list of documented impacts is substantial, less is known about the mechanics of wildlife/highway interactions. However, a review of literature pertaining directly to where wild animals interact with roads suggested that at-grade crossings do not occur at random locations and that variables from both the local and the landscape scale play a role in the location of wildlife/highway interactions.
To verify if the locations where animals cross highways are different from random locations, I recorded where wild animals crossed the road and measured characteristics of the surrounding habitat at two locations in the Southern Rocky Mountains. Depending on the type of data collected, I analyzed it by comparing average values, using a Monte Carlo approach to compare expected with actual distributions, or by directly comparing used with available habitat. I also collected data about underpass use and summarized it with simple counts.
A qualitative assessment of my results indicated that crossing zones are related to variables from both the landscape and the local scale. Significant variables
111


included features from both the habitat and roadway. At the large scale, the most important features were cover type composition, slope, and slope complexity of landscape surrounding the highway. At the local scale, the most important features were the location of and distance to roadside barriers, the location of drainages, and the distance from the road to the forest edge. These variables provide an accessible source of information to improve the practice of reducing wildlife/highway conflicts. Conflict reduction should include identifying conflict locations at both a landscape and a local scale, considering conflict locations when choosing the best strategies and locations for mitigation, and integrating this process into highway project planning up front. Additional data about wildlife/highway interactions as well as coordination among highway planners, conservation planners, and entities that affect land use, will improve the practice of reducing wildlife/highway conflict in the future.
This abstract accurately represents the contents of the candidates dissertation. I recommend its publication.
Signed
Willem K.T. Van Vliet-
IV


ACKNOWLEDGEMENTS
My thanks to everyone who helped make it possible for me to earn my Doctor of Philosophy. I extend special thanks to:
My advisor and all my committee members for their advise and support while I worked on this project.
The Colorado Department of Transportation for their generous funding and the freedom to conduct my research as I saw fit.
My field assistants, Eli Wostl, Christiana Manville, and Marcus Pacheco, who always worked hard and did a great job.


CONTENTS
Figures.......................................................xii
Tables........................................................xiv
CHAPTER
1. REDUCING WILDLIFE/HIGHWAY CONFLICTS..........................1
Introduction..............................................1
An Overview of Wildlife/Highway Conflicts.................2
Habitat Fragmentation..............................3
Habitat Loss and Alteration........................4
Direct Mortality...................................6
Disturbance........................................7
A Historical Perspective of Wildlife/Highway Conflicts....8
Traditional Highway Planning and Wildlife..........8
Traditional Conservation Planning and Highways....14
The Current Institutional Setting for Reducing Wildlife/Highway Conflicts................................................18
The Regulatory Nexus to Consider Wildlife in Highway Planning..........................................20
Highways and the Ecosystem Approach to Conservation Planning..........................................23
vi


Hie Current Practice of Reducing Wildlife/Highway Conflicts......................................25
Summary..............................................27
2. A REVIEW OF RESEARCH RELATED TO IDENTIFYING
WILDLIFE HIGHWAY-CROSSING ZONES.........................28
Introduction....................................... 28
Roadside Habitat Use.................................30
Animal/Vehicle Collision Locations Along Highways....31
Use of Highway-Crossing Structures by Wildlife.......33
Locations of At-Grade Highway Crossing by Wildlife...35
Summary.......................................... 38
3. IDENTIFYING WILDLIFE HIGHWAY CROSSING ZONES IN
THE COLORADO ROCKIES: RESEARCH BACKGROUND................39
Introduction.........................................39
Overview.......................................39
The Research Approach..........................40
Relevance of the Measured Variables................ 41
Landscape Scale Variables......................42
Local-Scale Variables........................ 44
Study Site Descriptions..............................45
Trout Creek Pass...............................46
VII


Vail Pass..........................................49
Data Collection Methods: Tracking.........................54
Standard Tracking Methods..........................54
Standard Underpass Monitoring......................56
Snow Tracking Methods..............................57
Animal Abundance...................................59
Data Collection Methods: Habitat Measurement..............60
Landscape Scale Habitat Measurements...............60
Local Scale Habitat Measurements...................62
Data Analysis: Identifying Patterns.......................64
Descriptive Summary of Track Records............. 65
Identifying First Order Patterns...................65
Identifying Second Order Patterns (Crossing Zones).65
Measuring Crossing Zones...........................67
Underpass Use......................................67
Data Analysis: Quantifying Relationships..................68
Quantifying the Relationship of First Order Patterns to Landscape Structure................................69
Quantifying the Relationship of Second Order Patterns to Local Scale Features...............................69
Quantifying Animal Abundance.......................71
Summary...................................................72
vui


4. WILDLIFE HIGHWAY-CROSSING ZONE INVESTIGATION: FINDINGS.........................................................73
Introduction..............................................73
Descriptive Summary of Tracks....................... 73
Distribution of Tracks................................ 75
First Order Patterns...............................75
Second Order Patterns..............................80
Crossing Zones.....................................81
The Relationship of First Order Patterns to Landscape-Scale Features..................................................86
Composition........................................89
Complexity..................................... 90
Other Influences from the Landscape Scale..........93
The Relationship of second Order Patterns to Local Scale Features................................................ 94
Features Measured Directly in the Field............94
Features Measured from GIS Data Layers.............96
Underpass Use.......................................... 103
Trout Creek Pass..................................104
Vail Pass....................................... 106
Vail Pass Snow....................................108
IX


Animal Abundance........................................110
Trout Creek Pass..................................110
Vail Pass.........................................Ill
Vail Pass Snow....................................Ill
5. VARIABLES THAT IDENTIFY WILDLIFE HIGHWAY CROSSING ZONES: DISCUSSION......................................113
Introduction............................................113
Descriptive Summary of Tracks and Their Distribution....114
First Order Patterns and Landscape-Scale Variables......115
Composition and Complexity........................115
Other Landscape Scale Influences..................117
The Relationship of Second Order Patterns to Local Scale Features.............................................. 120
Variables Measured Directly in the Field..........120
Features Measured from GIS Data Layers............122
Underpass Use...........................................126
Trout Creek Pass..................................126
Vail Pass.........................................127
Vail Pass Snow....................................127
Summary.................................................128
x


6. APPLYING THE RESEARCH RESULTS TO REDUCE
WILDLIFE/HIGHWAY CONFLICTS..................................129
Introduction..........................................129
Are the Results of the Study Useful?.............130
Strategies for Identifying Wildlife/Highway Conflict Locations.....................................131
A Strategy for Identifying Conflict Zones........131
A Strategy for Identifying Crossing Zones........135
Approaches for Reducing Wildlife/Highway Conflicts......137
Identify and Avoid High Conflict Locations.......138
Managing the Surrounding Landscape to Reduce Conflicts........................................139
Design Based Approaches for Reducing Conflicts.140
Integrating Mitigation Planning into Highway Planning...143
Improving the Practice of Reducing Wildlife/Highway Conflicts.......................................... 144
Summary.................................................147
APPENDIX
A. ABBREVIATIONS USED IN THE TEXT.............................148
BIBLOGRAPHY......................................................149
xi


FIGURES
Figure
3.1 The location of the Trout Creek Pass study site (TCP) in Chaffee
County, Colorado.....................................................47
3.2 Detail of the Trout Creek Pass area, location of TCP...............48
3.3 Location of the Vail Pass (VP) and Vail Pass Snow (VPS) study sites
straddling Eagle and Summit Counties in Colorado.....................50
3.4 Detail of the Vail Pass area, location of both VP and VPS, including the
location of the Copper Mountain Resort (CMR) ......................51
3.5 The locations of all barriers and underpasses on 1-70 at Vail Pass..53
4.1 Locations of all crossing TRs recorded at TCP........................77
4.2 Locations of all crossing TRs recorded at VP........................78
4.3 Locations of all crossing TRs recorded at VPS..................... 79
4.4 Locations of the crossing TRs that make up the CZs in the north end of
TCP....................................................... .........82
4.5 Locations of the crossing TRs that make up the CZs in the south end of
TCP..................................................................83
4.6 Locations of the crossing TRs that make up the CZs on the east side of
VP...................................................................84
4.7 Locations of the crossing TRs that make up the CZs on the west side of
VP............................................................... 85
4.8 Locations of the crossing TRs that make up the CZs in the sub-area of
VPS adjacent to CMR..................................................87
vi


4.9 Locations of the crossing TRs that make up the CZs in the sub-area of
VPS not adjacent to CMR.................................................88
4.10 Detail of the summer crossing TRs that made up CZs, in relationship to
barriers and underpasses, on the east side of VP......................98
4.11 Detail of the winter crossing TRs that made up CZs, in relationship to
barriers and underpasses, on the east side of VP......................99
4.12 Detail of the summer crossing TRs that made up CZs, in relationship to
barriers and underpasses, on the west side of VP......................100
4.13 Detail of the winter crossing TRs that made up CZs, in relationship to
barriers and underpasses, on the west side of VP......................101
6.1 A framework for mitigating wildlife/highway conflicts along existing
highways :..............................................................133
Vll


TABLES
Table
1.1 Environmental statutes with potentially extensive effects on the placement
and design of highways in the U.S...................................22
2.1 A summary of variables demonstrated to be important in determining the
locations where animals interact with roads and highways............29
3.1 Habitat and roadway variables measured to compare crossing zone (CZ)
locations to random locations.......................................42
3.2 Definitions used to divide cover, slope, and aspect classses on digital
maps of the three study sites.......................................61
3.3 Variables measured in the field at crossing zone (CZ) points and random
points for comparison of local scale habitat characterisitcs........63
4.1 Summary of TRs by species and travel at TCP....................*....74
4.2 Summary of TRs by species and travel at VP. .........................74
4.3 Summary of TRs by species and travel at VPS..........................74
4.4 Distribution of crossing TRs within the sub-areas of each study site.80
4.5 Comparisons of sub-area pairs by composition.........................90
4.6 Comparison of landscape metrics associated with first order patterns.91
4.7 Average values of measurements taken at CZ and random point locations... 95
4.8 Results of chi-square tests comparing the distribution of cover, slope, and
aspect classes within 100 m of the CZs to what is available throughout the highway corridor within 100 m of the roadside...................97
vm


4.9 Actual distances compared with the expected distances of CZs to barrier
ends and the results of the chi-square tests comparing the actual with the expected number of TRs located mid-barrier.........................102
4.10 Actual distances compared with expected distances of CZs to the
nearest drainage that intersects with the road..........................103
4.11 Relationship of CZs to the forest edge.................................103
4.12 Characteristics of monitored underpasses at TCP and the number of
times at least one through-pass by at least one medium- or large-sized mammal was recorded.............................................. 105
4.13 Number of times at least on individual of a species used each
underpass at TCP........................................................105
4.14 Characteristics of monitored underpasses at VP and the number of times
at least one through-pass by at least one medium- or large-sized mammal was recorded...............................................107
4.15 Number of times at least on individual of a species used each
underpass at VP.........................................................107
4.16 Characteristics of monitpred underpasses at VPS and the number of
times at least one through-pass by at least one medium- or large-sized mammal was recorded............................................. 109
4.17 Number of times at least on individual of a species used each
underpass at VPS.................................................... 109
4.18 The total number of animals that used the underpasses at VPS...........109
4.19 The number of animal trails/transect recorded in the snow along offroad transects at TCP and VPS, by species...............................Ill
5.1 Results of the Mann-Whitney test comparing means of local-scale habitat measurements taken at point locations within CZs and random point locations..................................................... 120
ix


CHAPTER 1
REDUCING WILDLIFE/HIGHWAY CONFLICTS Introduction
Roads arguably create humans most pervasive physical impact on landscapes throughout the world (Forman, 1998). Yet, despite their extensive character and wide array of negative effects on nature, biodiversity/road conflicts have only recently come under scrutiny (Hordequin, 2000). As the list of documented road-based impacts to wildlife species and their habitat expands, it is becoming apparent that highway and wildlife conservation planners must cooperate in order to reduce these impacts. However, it is equally apparent that additional information about highway/wildlife interactions is required to create a successful collaboration (Forman, 1998).
For example, highway and conservation planners who wish to reduce roadkill and/or road-caused habitat fragmentation need baseline information about how wild animals interact with roadway. Both these conflicts are linked to the ability of animals to cross highways. Therefore, information about preferred crossing locations would be immensely useful for designing highways that avoid or mitigate high conflict areas, and allow animals to move safely and freely across the roadway.
With the need to generate and apply this type of information in mind, the goal of my research was two-fold. First, I wanted to determine if the locations where wild animals cross highways are correlated to definable characteristics of the
1


surrounding environment and the roadway itself. Then, I was interested in how this type of information could inform the highway design process to reduce wildlife/highway conflicts. I hypothesized that:
Wild animals do not cross highways at random and the locations that they use can be correlated to features from the roadside and the surrounding habitat.
The identification of likely crossing locations is an accessible process that can be incorporated efficiently into highway design to reduce wildlife/highway conflicts.
To place the need to generate and apply this type of information in context, I begin this Chapter by reviewing what is known about of wildlife/highway conflicts. Next I give a broad overview of the development of both highway and conservation planning and the historical conditions that prevented these two sets of specialists from generating and using knowledge about wildlife/highway conflicts to reduce impacts. I conclude by examining the existing institutional settings that form a basis for highway and conservation professionals to join forces and create highways designed with wildlife in mind. I then briefly examine the current practice of mitigating highway impacts on wildlife.
An Overview of Wildlife/Hiehwav Conflicts
Impacts of modem highways on the natural environment in general and on wildlife populations in particular are extensive and relatively well documented (Bennet, 1991; Evink et al., 1996,1998; Spellerburg, 1998; Trombulak and Frissell, 2000). There are four broad categories of impacts that may occur to wildlife as a result of highway projects including 1) habitat fragmentation 2) habitat loss and alteration,
2


3) direct mortality, and 4) disturbance. These four impact types can take different forms and may be direct or indirect.
Habitat Fragmentation
Highways fragment habitats when they create a physical or behavioral barrier to animal movement. This barrier effect occurs when animals avoid habitat near roads, are physically unable to cross a road, or are killed while attempting to cross. The consequences of these effects are restrictions on daily movements between resource areas as well as on long distance dispersal and seasonal migration. These types of impacts may have significant effects at the population level. For example, Sweanor et al., (1999) studied metapopulation dynamics of cougars (Felis concolor) in New Mexico. They documented a loss of dispersal capability by cougars across an expanded highway. Because one side of the highway acts as a source habitat for a smaller block of habitat on the other side, they concluded that the cougar population in the smaller block of habitat is unlikely to persist over time.
The degree to which a barrier contributes to habitat fragmentation is scale dependent. For example, at a local scale, if a road passes through habitat that is not preferred by a species, there would appear to be no barrier effect. However, if that area lies between two areas of preferred habitat, the road may act as a barrier at the landscape scale. An animals perspective also dictates the magnitude of barrier effects. While most deer most probably easily cross a roadway 20 m in width, it may be an insurmountable barrier for small mammals or species that are behaviorally averse to entering open areas. No quantitative data exist regarding how a roads design regulates its barrier effect. However, it is logical to assume that barrier effects increase for all species with increased width and the addition of
3


retaining walls, fences, raised medians, guard rails, and significant increases in volume and/or speed of traffic.
Habitat Loss and Alteration
Direct habitat loss occurs when a highway projects footprint removes natural cover. In addition to effects that result from a simple loss of cover, the impact of habitat loss depends on the overall availability of a habitat type in the project area and its role in an animals life cycle. Loss of abundant habitat types may be relatively inconsequential; loss of a rare habitat type will have a greater impact, especially if it provides an essential resource for a species (e.g., nesting/denning sites, important food sources). Other impacts may cause loss of habitat indirectly by restricting access.
In addition to direct habitat loss, highways also alter the environment in other ways that can be detrimental to wildlife. Replacement of native cover types with non-native roadside plantings alters habitat, and cut and fill construction techniques obliterate natural landforms. The physical presence of highways, as well as the cuts and fills, can change both surface and ground water flows. Changes to water flows often cause secondary environmental alterations which in turn can cause major alterations to habitat. Depending on the location of a roadway in the landscape, it may cut some areas off from normal water flows while consequently directing higher peak flows and more intense floods to others. Increased water flows and the accompanying debris and sediments can cause physical alteration of terrestrial and aquatic environments at both the local and landscape scale (Jones et al., 2000).
Highways are also widely documented to have significant impacts on water quality due to inputs of heavy metals, salts, and nutrients associated with the roadway
4


(Trombulak and Frissell. 2000). Changes in hydrological regime and water quality can cause changes in plant dispersal and survival, leading to plant community shifts over time. For example, Findlay and Bourdages (2000) found significant correlations between lower plant diversity and higher densities of roads within Canadian wetlands. Because these types of disruptions affect basic ecological processes, highway impacts may extend far beyond the roadside, affecting an entire ecosystems function (Forman, 2000).
Another impact of highways that can cause profound alterations to natural communities is the role they play in aiding the spread of non-native species and expanding the range of native species into previously unoccupied habitats. Both plant and animal species take advantage of the new pathways created by roads, and these invasive species often have deleterious effects on resident species that are not adapted to competing with them. Plants species may be spread intentionally as roadside plantings, or unintentional by seeds contained in mulches and erosion-control coverings. Additionally, the disturbed soils and sunny, treeless roadsides of newly constructed road projects create ideal growing conditions for most pioneer species, many of which are adapted to disperse widely via wind, water, or animal carriers. Animals may follow along roads during dispersal because the artificially smoothed roadside offers easy travel, or they may be lured along roadsides by new food resources created by the roadside plant community. This resource includes the plants themselves, or increased populations of small mammals and birds responding to the new cover type, which in turn attracts predators. Additionally, human users of the road and adjacent areas often leave behind trash that can also be an attractive food source for some animal species.
5


Direct Mortality
Although no data exist quantifying the extent of construction-related wildlife mortality (Trombulak and Frissell, 2000), it is logical to assume that small, local populations of sessile or slow-moving organisms may suffer significant negative consequences from construction. Data documenting mortality due to daily operations of highways, i.e., collisions with vehicles, do exist and cover a wide range species, including mammals, birds, amphibians, and invertebrates (Trombulak and Frissell, 2000). Roadkill is known to be a limiting factor for some populations of endangered species, including the American crocodile (Crocodilia americanus; Kushlan, 1988), and the Florida panther (Foster and Humphrey, 1995). Roadkill is also known to have devastating local effects on small populations. For example, Jones (2000) reported that a road upgrade in Cradle Mountain-Lake St Claire National Park, Tasmania, promoted higher vehicle speeds, resulting in increased road-kill rates for eastern quo 11s (Dasyurus viverrinus) and Tasmanian devils (Sarcophilus laniarius), both of which were subsequently extirpated from the park as a result of this increased mortality.
However, although an estimated one million vertebrates are killed daily on Americas roads (Defenders, 2002), accurate, current data about the mortality rate and population-level impact on common species are largely unavailable. For example, an estimated minimum one-half to three quarter million deer are annually killed nationwide (Romm and Bissonette, 1996a; Hubbard et al., 2000), yet there have been few efforts to document the actual number or rate of deer mortality as a result of collisions with vehicles (Romin and Bissonette, 1996a).
6


Disturbance
Disturbance results from many sources, including construction, day-to-day road operations, and increased human access to the area as a result of road improvements. Possible impacts of disturbance include direct mortality, temporary avoidance of an area, and permanent abandonment of the surrounding habitat. All these impacts may potentially interrupt activities (e.g., feeding, breeding, travel) essential to survival at both an individual and a species level. Additionally, disturbance may contribute to both habitat loss and fragmentation.
The impacts of construction-related disturbance are a function of the species susceptibility to disturbance, duration of the disturbance, area affected, type of disturbance (e.g., heavy equipment noise versus blasting noise), season, and time of day. Disturbances that last a long time, are loud, unpredictable, and/or affect large areas will have the greatest impact. Day-to-day road operations have been shown to cause permanent disturbance effects. Many species are known to avoid areas of disturbance, thereby reducing or eliminating the habitat value of these areas. Types of disturbance from highway operations include noise, visual stimuli, human activity, and pollution. Research (van der Zande et al., 1980; Reijnen et al., 1995; Reijnen et al., 1996) indicates that breeding bird densities are reduced near roads, with the effect being greater for heavy traffic and reaching farther in open habitats (up to 2000 meters) as compared with forested habitats (up to 1500 meters). In Colorado, both mule deer (Odocoileus hemionus) and elk (Cervus elaphus) were shown to avoid areas within 200 meters of a road, with this effect appearing stronger in shrub cover types, as compared with forested habitats (Rost and Daily 1979). Studies also indicate that a variety of carnivores, including grizzly bears (Ursus horribilis; McLellan and Shackleton, 1988), wolves (Canis lupus; Thiel, 1985; Mech et al., 1988) and bobcats (Lynx rufus; Lovallo and Anderson, 1996)
7


avoid habitats adjacent to roads. Permanent disturbance effects will contribute to both habitat loss and fragmentation.
Road projects also often provide increased human access to previously unused areas. Increased human activity can severely reduce or eliminate the habitat value of an area for many species by eliciting an avoidance response. Shy species, such as most felids, are particularly susceptible to this impact. Increased human presence will also contribute to habitat loss and fragmentation.
A Historical Perspective of Wildlife/Highwav Conflicts
The histories of highway planning and wildlife conservation planning in America share a similar time line. Both disciplines began informally upon the settlement of North America by Europeans, became more formalized in the early 1900s in response to increasing scientific knowledge and public pressures, and then continued to evolve under these two influences from the 1950s onwards. Both disciplines also have largely ignored one another until very recently. Although highways can have profound effects on wildlife, as discussed above, most planned highways have been designed with almost no consideration for the environment, while conservation theorists and planners made few, if any attempts to comprehensively address these impacts, offer solutions, or otherwise engage highway professionals. I discuss the history of these two disciplines in the U.S. below.
Traditional Highway Planning and Wildlife
From European settlement until the early 1900s, American highways were not planned, per se. The natural environment, in the form of the landscape and its hydrological and geological processes, and the needs of local human populations
8


dictated the location and form of highways in America. These highways usually did not connect point A to point B by the most direct route. Instead, they followed the line of least resistance through the intervening landscape (Lane, 1950). Early highway construction consisted simply of clearing vegetation, and perhaps limited grading, filling of wet areas, and constructing bridges across small streams (Patton, 1986). As a result, the environmental impacts of these highways, including impacts to wildlife, were low. These highways did not constitute major barriers to animals. Habitats were not fragmented except at the most local scales. Landforms were not altered to accommodate the road. Sensitive habitats that were difficult to traverse, such as wetlands and high elevation zones, were avoided. Low traffic speeds meant that roadkill was only an issue for the smallest, slowest moving animals.
Beginning in about 1890, the Good Roads Movement blossomed in America. Bicycling clubs, which wanted improved surfaces away from crowded local roads for their new form of recreation, were the main proponents for improved roads initially (Patton, 1986; Kaszynski, 2000), but about this time the automobile appeared on the scene as well. By 1902 over 50 auto clubs had formed nationwide and joined the call for Good Roads (Kaszynski, 2000). In response to these public demands, improved highways were constructed using newly codified engineering principles including surveying, siting, and surfacing techniques, as well as application of geometric design to the shape of the roadbed (USDOT, 1976). The application of these principles can also be considered Americas first widespread use of formal highway planning, and it did not include any considerations to reduce impacts to the natural environment. In the early 1900s it was not yet apparent to anyone that roads could cause negative impacts to natural systems. The formal study of ecology was still in its infancy (Kingsland, 1991) and understanding of ecological function was consequently almost non-existent at this time. Roads had
9


previously caused only the most minimal impacts, and the natural environment was not yet widely viewed as a resource in need of protection.
Initially, the form and location of these planned highways did not differ much from traditional rights-of-way, and the alignment and grade remained conditioned on horse drawn traffic (Hewes, 1950). By 1925, however, automobile ownership was common, and highway engineers realized that the consequent increasing average traffic speed called for additional innovations in highway design. Speed-friendly designs required moderate grades and gentle curves qualities that often did not occur along the path of least resistance which most traditional rights-of-way followed through the landscapes. The solution to this problem was to reshape the landscape to fit the needs of the highway. By 1940, standard highway designs reflected an increasing tendency among highway engineers to break away from conditions previously accepted as determining factors in highway location (Hewes 1950, p. 327).
As the form of the highway departed from the form of the landscape, and traffic speeds increased, roadkill became a noticeable phenomenon, as evidenced by three early journal articles. The Toll of the Automobile (Stoner, 1925), Birds and Motor Cars (Cottam, 1931), and The Automobile as a Destroyer of Life (Davis, 1934) are probably the first published accounts of environmental impacts associated with highways. However, these authors focused on the impacts of using the highway, and probably had a very limited audience. Widespread appreciation of environmental impacts caused by highway itself remained elusive at this point, with one exception. This was the visual and aesthetic experience of driving on the highway.
10


The push to create an efficient American highway system during the early 1900s had focused limited funds on pavement rather than the roadside (Neale, 1950). However, as traffic volumes increased along urban highways, a clutter of billboard advertisements and businesses sprang up along them, creating both a safety hazard and a visual blight. The public, accustomed to natural country roads, also found new, unlandscaped rural highways ugly and maintenance issues associated with these untreated roadsides were a significant problem. The bare, erosion-prone cuts and fills which many road projects created made road washouts commonplace (Robinson, 1971).
Partly in response to these concerns the American Association of State Highway Officials (AASHO) and the Highway Research Board (HRB) joined forces in 1930 to form a committee to study all phases of roadside design construction, and maintenance. When the Joint Committee disbanded in 1940, AASHOs Committee on Roadside Development took up its work. It focused its attention on the basic principles of landscape design and practice which should be incorporated in postwar plans for all highway construction. Its findings were summarized in a 1943 report that advocated construction of the complete highway that would be designed to maximize the four basic qualities of utility, safety, beauty, and economy (Neale, 1950, p. 321). The report was, however, merely a recommendation and was never adopted as official policy by either AASHO or the Bureau of Public Roads. Highway project planners were free to interpret or ignore the four basic qualities as they saw fit.
Nevertheless, the quality of beauty in effect became the yardstick whereby environmental impacts were acknowledged and measured both by highway engineers and highway users. For this assessment, only the environment directly adjacent to the roadside was of interest. During the 1940s, the science of ecology
11


was still in its infancy and the concept of ecosystems and their interconnections were not yet widely disseminated (Kingsland, 1991). Very few people understood that a narrow, linear structure like a highway could have effects on natural systems that extended far beyond the roadside. It was assumed that all the impacts of the highway could simply be constructed away with artful grading and naturalistic plantings that blended with the surrounding landscape.
This very limited interpretation of the impacts of highways on natural systems was inadvertently reinforced by the 1956 Highway Act (and all subsequent Highway Acts until the 1990s) that funded construction of the interstate highway system. The Act specified that planners and designers had to follow AASHOs geometric design standards in order to receive funding for their project. These standards had been adopted as part of the Act to ensure highway consistency nationwide (USDOT, 1976). This single design mandate effectively narrowed the focus of highway designers. AASHOs design standards were based on the proposed design speed and projected traffic volume only. No other considerations were mandated, although they were also not forbidden. In practice however, the AASHO standards were interpreted to mean that only design speed and traffic volume could be considered, and state highway agencies began planning and designing to those criteria, to the exclusion of all other considerations (Myerson, 2000). Roadside beautification projects continued to be incorporated in many projects, but any concern about impacts beyond the immediate roadside was effectively short-circuited by these narrow standards.
The pavement only emphasis created by the Highway Act funding was further compounded by the elitist attitudes of highway experts and the great pressure they faced from the public and private sectors to get the interstate built. The Good Roads era of the early 1900s had seen the creation of the American highway professional.
12


Congress intentionally appointed an engineer to lead the first federal road agency, in part because road building was becoming increasingly technical, and in part because it wanted an aura of apolitical expertise associated with federally funded highway projects (Seely, 1987). Thus, from the beginning, highway professionals had both a self- and a public-image of impartial experts applying a complex science that only they understood. By the 1950s, this attitude was deeply entrenched. The engineers and bureaucrats who controlled the highway building process did not consider contributions from non-engineers valid (Patton 1986), and they actively disregarded input from the public and specialists from other disciplines in regards to highway projects.
Additionally, by the time construction on the interstate system was initiated in 1956, the existing highway system, constructed mainly before World War II, was hopelessly overburdened. Faced with an overwhelming clamor for better roads, from both the public and commercial interests, the main priority of the state agency highway professionals charged with building the interstate was to construct efficient roadways that maximized user benefits, as quickly as possible (Kaszynski, 2000). The push for efficiency, defined as more cars, faster and safer, inspired direct alignments from point A to point B. In order to achieve this goal, curves of the landscape were cut, wetlands were filled, streams channalized, and remote mountain passes were paved. Cuts and fills, combined with wider roadways, unnatural surfaces, and higher traffic speeds and volumes created substantial barriers to wildlife movements, and the silt-laden run-off from roadsides had severe impacts to aquatic systems (Mowbray, 1969). However, the impact of all these habitat alterations were not perceived by highway planners and designers as important, if they were perceived at all.
13


Throughout the 1960s impacts to natural systems from highways and a myriad of other sources continued to grow. In response, the environmental movement in America was bom and Congress was pressured into passing legislation to protect the environment. I review statutes that pertain to highways in detail in the next section of this chapter. However, even with these statutes in place, the attitudes of highway professionals and the design standards required to receive highway funding continued to make it difficult to design highways to anything but the needs of the roadway itself. Although other considerations were not forbidden, safety and efficiency at the posted speed continued to be the only criteria recognized in AASHTOs design standards. Because the vast majority of highway professionals did not share the publics new-found environmental awareness (Mowbray, 1969; Lewis, 1997) and because of their deeply entrenched negative attitudes about the input from non-engineers, narrow interpretation of these standards continued. Most highway projects through the 1970s and 1980s were designed to the roadways needs and then minimally modified to accommodate other regulatory requirements (Wick, 1995).
Traditional Conservation Planning and Highways
As reviewed below, conservation planning in the U.S. has a long history of ignoring how conservation goals might be attained in context of the built environment. While this outlook certainly includes a lack of consideration about how highways impede conservation goals, as well as a lack of interest in how highway planning and design might by modified to minimize these impacts, it would be incorrect to claim that highways have been singled out for this attitude. Two strong themes in Western culture, the tendency to see nature as something separate and alien from humans, and a tendency to place nature on a pedestal have
14


created a strong schism between planning for conservation and planning for all aspects of the built environment.
Actions that can be construed as wildlife conservation planning began in the American Colonies shortly after settlement with the passage of laws specifying the hunting season for deer in Rhode Island, and continued during the 1700s and 1800s with the passage of similar statutes throughout the country (Leopold, 1933; Andrews, 1999). During this entire period, wildlife was treated separately from the management of other natural resources such lumber and water, and there was no attempt to account for wildlife values as agricultural clearing, settlement, and road building occurred. However, by the mid-1880s it was becoming apparent that animal populations were not unlimited, and that additional management would be required to prevent their eventual destruction (USDI, 2001). The advent of recreational (as opposed to subsistence) hunting and wildlife observation as a hobby also began in the late 1800s, and this public interest in multiple uses for wildlife required a different approach to management as well.
At the turn of the 20th century, President Roosevelt and Forest Service head Gifford Pinchot began to champion an approach to natural resource management that they termed conservation through wise use. This movement sprang from a belief that resources needed to be conserved in order to provide future benefits (raw materials, food, amusement) to humans, but it represents the first widespread linking of the term conservation to the stewardship of natural resources. The movement also had two other important principles, the recognition that all natural resources comprised an integrated whole, and the recognition of science as an important tool for informing natural resource management. This interest in the scientification of natural resource management, combined with the newly applied idea of
15


conservation, helped to create the field of resource management. This new class of experts was also, in effect, Americas first professional conservation planners.
However, although these experts increasingly thought of nature as a whole, they did not break from the traditional Western cultural view that nature and humans are separate (Western, 1989). This view was expressed in creation of the first National Parks in the late 1800s and the first wildlife refuges in the early 1900s, which focused the practice of wildlife conservation on setting patches of habitat aside from humans. Despite the vast growth in the knowledge and theory that supports the practice of conservation planning, discussed below, this mind-set remained embedded in conservation planning throughout the 20th century due to both philosophical outlooks and for practical reasons. The concept of delineating and managing a separate patch for wildlife fits in well with both western cultural norms and with normal sciences orientation towards simple, orderly solutions based on single, knowable truths (Lister, 1998). Practically, planning and managing a wildlife refuge, a single-use patch with definite boundaries, is a simple task, as compared to the wicked problem (Rittle and Webber, 1973) of planning and managing an ill-defined area for multiple uses.
Concurrent with the adoption of a scientific approach to planning for and managing wildlife was the growth of the science of ecology which provided an additional source of expert information for natural resource mangers. The Ecological Society of America was officially constituted in 1915, and the study of ecology flourished from the 1920s onward. As the field grew, an understanding of humans place in nature and the profound effects of human actions on natural systems emerged. Popular works by Aldo Leopold (1949) and Rachel Carson (1951,1955,1962) made these new ecological concepts accessible to non-scientists. However, the widespread recognition of humans as an integral part of the natural world did not
16


translate into an integration of the management of human infrastructure and natural resources. Although the birth of the environmental movement in the 1960s began to break down the long held societal norm that nature was separate and alien form human endeavor, it was replaced to some extent with the notion that humans were so damaging to nature that it had to be kept separate in order to survive.
Thus, no matter to what school of thought wildlife conservation planners subscribed, there was an underlying tendency to ignore the built environment that increasingly penetrated and surrounded wildlife habitats. This outlook was probably compounded by their professional training, which quite naturally focused them on their own field of expertise. The ecological sub-discipline of reserve site selection offers a clear illustration of this single-mindedness. This area of theory, regarding the best size and arrangement of reserves, as well as the best schemes for connecting them (e.g., Diamond, 1975; Pickett and Thompson, 1978; Soule and Simberloff, 1986; Simberloff and Cox, 1987) was developed by ecologists as a direct result of their interest in conservation planning. It was built largely upon the theory of island biogeography (MacArthur and Wilson, 1967), which had been in turn developed from basic ecological principles, that by definition are not concerned with influences of the human environment. Therefore, it is not surprising that most reserve site theories focused strictly on natural constraints, and did not account for the human land-uses and infrastructure that might already be in place. Additionally, little interest was accorded as to how areas surrounding reserves, unprotected and subject to multiple uses including highway construction, might be planned for, designed, or managed to increase its value to wildlife (Prendergast et al., 1999).
During the 1980s, the reserve configuration discussion and its interplay with attendant topics such as genetics and population biology helped give rise to the
17


field of conservation biology (Soule, 1985). Its practitioners dominate the field of conservation planning today. Although this latest class of conservation planners does not unerringly heed its roots, conservation biologists continue to frame much of their effort in terms of reserve creation (e.g., Noss, 1992; Soule and Terborgh, 1999; Hoctor et al., 2000; Margules and Pressey, 2000). Although most authors acknowledge that meeting conservation goals may be difficult due to competing land uses, only rarely do they examine the constraints this reality places on a reserve-centric approach to conservation. Likewise, there has been little emphasis on developing approaches that focus on meeting conservation goals in areas that have not been set aside as reserves (Prendergast et al., 1999). Among wildlife experts, the concept of conservation planning remains largely a biologically based process, conducted independently of planning for other activities and projects that also affect land use.
Despite the narrow focus of many conservation planners, finding ways to create and maintain habitat that can sustain wildlife while subject to other uses has received at least theoretical attention in other fields, notably land-use planning and landscape ecology (e.g., McHarg, 1969; Western, 1989; Soule, 1991; Hansen et al., 1991; Hansen et al., 1993; Shafer, 1994; Dale et al., 2000). Never the less, efforts to integrate wildlife considerations into planning for the built environment are by no means standard procedure. However, some important first steps to integrate conservation planning into highway planning and design have been taken and are discussed below.
The Current Institutional Setting for Reducing Wildlife/Highwav Conflicts
The problem of highway-caused impacts to the natural environment is becoming more widely acknowledged by highway professionals. For example, the
18


Transportation Research Board established the Committee for Environmental Analysis in Transportation in 1991, and sponsored workshops at its annual conference in both 2001 and 2002 to define the concept of environmental stewardship and create a framework for state transportation departments to institute it. AASHTO created a Standing Committee on the Environment in the early 199Qs and highlighted environmental stewardship as one of the three focal topics at its 2002 annual meeting.
Likewise, there is an increasing recognition of highway-related issues in the conservation community. Academics, agency personnel, and environmental groups are studying and discussing conflicts between highways and the natural environment. Examples of the growing interest in the topic include the special section on the ecological effects of roads in issue 1, volume 14 (2000) of Conservation Biology, Richard Formans new text Road Ecology (2002), and the 2001 International Conference on Ecology and Transportation which was attended by over 340 individuals and sponsored by 11 organizations, including the U.S. Forest Service, U.S. Fish and Wildlife Service, the Wetlands Division of the Environmental Protection Agency, Defenders of Wildlife and the Humane Society of the United States.
There is no single reason that the negative impacts of highways and wildlife have lately become a topic of concern. An important general factor is probably the increasing awareness among many sectors of American society that we live in a world of finite resources. Two specific manifestations of this cultural change playing a role in raising the profile of highway/wildlife conflicts are: 1) the regulatory mandates that currently control federal funding of highway projects and 2) the advent of the ecosystem approach in natural resource management. Both of these themes are shaping the current institutional setting that is encouraging
19


conservations professionals to press their concerns about highways and highway professionals to address wildlife/highway conflicts as part of project planning and design.
The Regulatory Nexus to Consider Wildlife in Highway Planning
In the mid 1960s, Congress began passing legislation to force highway departments to plan and design highways that would ameliorate negative impacts. Regulations directly applicable to wildlife issues included the Department of Transportation Act of 1966, which forbade the use of federal dollars on projects that would adversely impact publicly owned land that was used as a park, recreation area, or wildlife refuge, unless there is no feasible and prudent alternative to the use of such land [23 U.S.C., 138 section (4)(f)], and the 1970 Federal-Aid Highway Act which set noise standards, required project planners to coordinate with state and local air quality plans to meet standards set forth by the Clean Air Act of 1970, and mandated that the economic, social, and environmental effects of a highway project be considered before it could go forward.
During the 1960s and early 1970s, Congress also passed a number of laws specifically regulating impacts to the environment, which in turn could be applied to reduce highway impacts. The regulation with the most direct application to wildlife is the Endangered Species Act (ESA). However, statutes that regulate land use can also effectively prevent impacts to wildlife. Depending on the location of a highway project, a variety of environmental regulations can influence its placement and design, to the benefit of local wildlife. I list statutes that potentially have an extensive effect in Table 1.1. Each of these acts requires that adverse impacts to the resources they address be avoided or minimized. Conscientious application of these
20


regulations can have a significant effect on the level of impact created by a highway project.
However, generic environmental regulations do not require highway specific mitigation, and highway planners and designers have historically tended to ignore or only minimally complied with their mandates. Without a genuine commitment to the intent of these laws among highway experts, the success of these statutes often depends on the amount of external pressure a regulations champion applies on the highway project planning process. As a result of the uneven application of these statutes, public dissatisfaction with the process of highway planning and design as well as the outcomes of highway projects continued throughout the 1980s.
In response, Congress strengthened the basis for minimizing the impacts of highways to the environment through additional legislation. The Intermodal Surface Transportation Efficiency Act (ISTEA), passed in 1991, was designed mainly to reduce transportations impact to the human environment, but applied equally to the natural environment. In particular, ISTEA explicitly repealed the use of the AASHTO standards as a prerequisite for project funding and required that 10 percent of funds allocated for surface transportation be set aside for non-pavement transportation enhancement (TE) projects. Language within the Act emphasized preserving and protecting environmental and cultural values affected by transportation facilities (FHWA 1997, p. v). This created the first legal mandate for reducing environmental impact through sensitive project planning and design.
21


Table 1.1 Environmental statutes with potentially extensive effects on the placement and design of highways in the U.S.
Regulation Name Year Passed Influence on Highways
Wilderness Act 1964 Forbids any type of development on lands designated as wilderness by congress.
Highway Beautification Act 1965 Controls outdoor advertising and placement of junkyards adjacent to interstate and primary highway systems. Provides funding for roadside landscaping.
Wild and Scenic Rivers Act 1968 Forbids certain impacts to rivers designated as wild and scenic by congress.
National Environmental Policy Act 1969 Requires disclosure of environmental impacts caused by all federally funded actions, including highway projects.
Clean Air Act, Transportation Conformity Rule 1970 Requires that transportation projects conform to State air quality implementation plans.
Clean Water Act 1972 Requires the restoration and preservation of chemical, physical, and biological integrity of the nations waters through prevention, reduction and elimination of pollution from all sources. Forbids impacts to jurisdictional wetlands.
Costal Zone Management Act 1972 Preserve, protect, restore, and enhance resources of the coastal
Endangered Species Act 1973 Forbids negative impacts as a result of any federally funded project to fish, wildlife, and plant species designated as threatened or endangered.
Coastal Barrier Resources Act 1982 Minimize damage to fish, wildlife, and other natural resources due to projects occurring within the boundaries of a designated coastal barrier unit
During the mid and late 1990s, additional highway specific legislation expanded the vision ofISTEA and specifically emphasized reducing impacts to wildlife. The 1995 National Highway System Designation Act (NHS) specifically states that highway design may be tailored to the natural environment and to reducing environmental impacts. Although the statute is permissive rather than mandatory, it provides additional legal support for making the environment a primary consideration in highway design. Then, in 1998, the Transportation Equity Act for the 21st Century (TEA-21) expanded the list of TE activities established by ISTEA to include projects to reduce vehicle-caused wildlife mortality while maintaining habitat connectivity [23 U.S.C., section 101(a)(35)].
22


Another important development that helped to create a regulatory nexus for reducing highway impacts to wildlife was the FHWAs adoption of the planning model known as Context-Sensitive Design (CSD) in 1997. Responding to the mandates of NHS and TEA-21, the FHWA officially approved CSD as a tool to make considerations of both human and the natural environment a focal point in the design of highway projects. Although this guidance to the states was not legally binding, the policy document Flexibility in Highway Design legitimatized the concept of allowing a roadways surroundings to guide its form. This document, written as a companion volume to AASHTGs book of geometric design standards advocated (FHWA, 2001):
- Projects that satisfy a full range of stakeholders
- Stakeholder agreements forged early in the project and amended as needed
- Projects that provide safety for both the user and the community
- Projects that are in harmony with the environment
- Projects that are built with minimal disruption to the environment
- Projects that are seen as having lasting value to the environment
Highways and the Ecosystem Approach to Conservation Planning
The growing interest among conservation planners in addressing wildlife/highway conflicts is in part a spillover effect from the legal mandates affecting highway planning. However, an ongoing paradigm shift in the practice of wildlife management also plays a strong role. Wildlife conservation planning in the United States traditionally focused on creating and protecting reserve areas. When the human population was small, the surrounding matrix of unprotected area was largely undisturbed and augmented Americas relatively small total reserved area. However, as human populations increase, they place greater pressures on this
23


matrix area, and its ability to provide conservation values decreases, threatening the function of many reserves.
As a consequence, wildlife professionals are recognizing that simply creating preserves and managing them as natural islands, separate from the sea of development that is occurring around them, is unlikely to achieve conservation goals. Instead, these professionals realize they must explicitly acknowledge and incorporate the built environment, such as highways, into their planning to secure success. The spread of this new attitude is illustrated by the advent of ecosystem approaches for natural resources planning and management, beginning in the early 1990s.
The ecosystem approach is generally acknowledged as being a concept, rather than a prescriptive methodology for achieving a particular goal (Grumbine, 1994; Bom and Sonzogni, 1995; Brussard et al., 1998; Yaffee 1998). The practical application of an ecosystem approach varies, and it may be referred to as integrated environmental management, watershed management, or ecosystem management, among other similar names. The common threads of planning and management practices that fall under the ecosystem rubric are reliance on data about the system of interest, utilization of adaptive management, promotion of collaboration between different disciplines, agencies, and stakeholder groups, and acknowledgment of the role of values as well as science when setting and implementing conservation goals. Additionally, ecosystem approaches acknowledge ecosystem complexity and dynamism, multiple temporal and spatial scales, and ecological, as opposed to political, boundaries (Slocombe, 1993; Grumbine, 1993; Bom and Sonzogni, 1995; Christensen et al., 1996; Haeuber, 1996; Brussard et al., 1998; Yaffee, 1999).
24


Examples of the paradigm shift in the conservation planning community include adoption of an ecosystem approach by the federal government during the 1990s. The U.S Forest Service approved ecosystem management as the basis for natural resource management in 1992. A similar approach was subsequently institutionalized throughout the federal government in a Memorandum of Understanding (1995) that was signed by 14 federal agencies. The Nature Conservancy (TNC) provides another example. Historically, TNC was dedicated to simply purchasing and preserving reserve areas, but expanded its mission in 1996 to include management of landscape-level processes through an ecosystem approach (Poiani, 1996).
Two themes from the ecosystem approach to conservation planning encourage resource managers to consider the effects of the built environment, including wildlife/highway conflicts. The emphasis on recognizing ecological boundaries instead of the artificial boundaries of management areas promotes conservation professionals to acknowledge the impacts of infrastructure such as highways on both the protected areas and wildlife populations that move in and out of those areas. The emphasis on interdisciplinary collaboration sets the stage for interactions with highway planners and designers in order to find collaborative solutions that can reduce impacts.
The Current Practice of Reducing Wildlife Highway Conflicts
As discussed above, the conditions under which highway professionals plan and design highway projects have changed substantially since the 1960s, and especially during the past decade. Additionally, changing paradigms in natural resource management mirror this growing mandate for highway professions to include non-pavement considerations in their work. However, the actual practice of
25


reducing wildlife/highway conflicts remains in its infancy, and the procedures for incorporating mitigation into highway projects are haphazard. Most decisionmaking related to this type of conflict mitigation relies on expert opinion rather than data, and there are no objective standards in either the highway or the conservation community to determine if mitigation is sufficient.
Efficient implementation of projects designed to reduce conflicts is impeded by a lack of information about the nature of wildlife/highway interactions. Despite the relatively extensive literature documenting impacts of roads on the environment (reviewed above and in: Bennett, 1991; Spellerberg, 1998; Trombulak and Frissell, 2000), the ecological effects of roads have yet to receive attention as a unified category of research or conservation planning by ecologists and conservation biologists (Hourdequin, 2000). In particular, there is especially little data about how wild animals interact with the roadway as well as where these interactions are most likely to occur.
This lack of good information affects the ability of highway professionals to confidently plan mitigation projects. Roadway designers can reduce wildlife/highway conflicts by modifying the highway itself, for example by including wildlife underpasses, or they might choose alignments that avoid sensitive wildlife areas all together. However, because these types of design modifications are costly, the amount of mitigation that can be included may be limited by the project budget. Deciding how to allocate limited funding to do the most good is difficult without sound information.
The willingness of wildlife professionals to participate in wildlife/highway conflict mitigation planning is also impeded by the lack of information about wildlife/highway interactions. A hallmark of the ecosystem approach to natural
26


resource management is the reliance on data to create management strategies for the system of interest. Conservation planners increasingly demand good information in order to formulate their own plans or to buy into conservation oriented strategies designed by others. The unwillingness of some wildlife professionals, especially the regulatory personnel, to endorse mitigation projects proposed by highway professionals, is a strong disincentive for highway designers to include wildlife considerations in their plans.
Summary
Negative impacts of highways on wildlife include highway-induced habitat loss and alterations, habitat fragmentation, direct mortality, and disturbance. Although the list of documented impacts is substantial, less is known about the mechanics of wildlife/highway interactions. This type of basic information is needed to mitigate existing impacts and avoid future impacts. As the historical barriers between highway and conservation planners break down due to legal mandates and the advent of the ecosystem approach to conservation planning, the need for sound information to inform decision-making is growing.
27


CHAPTER 2
A REVIEW OF RESEARCH RELATED TO IDENTIFYING WILDLIFE HIGHWAYCROSSING ZONES
Introduction
Most of the literature documenting interactions of wildlife and roads (reviewed in: Bennett 1991, Spellerberg 1998, Trombulak and Frissell 2000) focuses on simply quantifying the negative effects of roads on wildlife and their habitats. To date, researchers have spent relatively little energy investigating the variables that regulate how wild animals interact with roads and that might be used to predict the location and severity of negative impacts. In particular, few studies have attempted to determine if there are specific locations where animals are more likely to cross highways, and to identify the characteristics associated with these locations. I review the small amount of existing literature that pertains directly to these two questions in this chapter.
Additionally, I examine literature regarding roadside habitat use, the use of crossing structures by wildlife, and the locations of animal/vehicle collisions (AVC). Because these types of studies also investigate where wild animals interact with roadways, they provide useful insight into the variables that are likely to be associated with the locations where animals cross the roadway. The research that I review is summarized in Table 2.1, then examined in detail below. I begin by covering the three attendant topics, and close by summarizing the existing research
28


Table 2.1 A summary of variables demonstrated to be important in determining the locations where animals interact with roads and highways
Author Research Type Important Variables Summary of Results
Alexander and Waters, 2000 Cross Topography Crossing associated with flat slopes, south-to-west exposures, low topographic complexity
Allen and McCullough, 1976 AVC Food* AVCs occurred near cover types providing food
Bashore et al., 1985 AVC Food AVCs occurred near cover types providing food
Beilis and Graves, 1971 AVC Food* AVCs occurred near cover types providing food
Carbaugh et al., 1975 Cross, Hab Roadside characteristics*, Food Roadside habitat uses associated with cover types providing food, crossings avoided roadside barriers, steep cuts & fills
Clevenger et al., 2001 Struc Roadway characteristics For most species structure use increased with traffic volume
Clevenger and Waltho, 2000 Struc Human activity Some species only use structures with low levels of activity
Feldhamer et al., 1986 Hab Food, AVCs occurred near cover types providing food
Finder et al., 1999 AVC Cover, Topography AVCs occurred near forest cover, drainages
Gibeau et al., 2002 Hab Habitat, Human activity Roadside use associated with preferred habit, low levels of human activity
Haas, 2000 Cross Cover, Habitat fragmentation Structure use associated with unfragmented habitats and cover near the entrance
Hubbard et al., 2000 AVC Topography, Habitat fragmentation AVCs associated with large cover-type blocks, drainages
Inbar and Mayer, 1999 AVC Habitat Roadside use associated with preferred habit
Lovallo and Anderson, 1996 Hab Habitat Roadside use associated with preferred habit
Lyon, 1979 Hab Cover Roadside use associated with forest cover types
Mamalis, 1995 Cross Roadside characteristics* Crossing avoided roadside barriers
Puglisi et al., 1974 AVC Food* AVCs occurred near cover types providing food
Reilly and Green, 1974 AVC Food* AVCs occurred near cover types providing food
Rodriguez et al., 1996 Struc Cover Structure use associated with cover near the entrance
Romin and Bissonette, 1996 AVC Topography, Cover AVCs associated with drainages, non agricultural cover types
Rost and Baily, 1979 Hab Cover Roadside use associated with forest cover types
Singleton and Lehmkuhl, 2000 Cross Habitat fragmentation*, Human activity* Roadside use associated with unffagmented habitats, low levels of human activity
Yanes et al., 1995 Struc Cover Structure use associated with cover near the entrance


directly concerned with identifying actual crossing-locations and quantifying their characteristics.
Roadside Habitat Use
A widely documented effect of roads on wildlife is the avoidance of areas adjacent to the roadway (Bennett 1991, Spellerberg 1998, Trombulak and Frissell 2000). However, some studies indicate that certain habitat characteristics appear to mediate this response. It is important to consider the results of studies that document this effect. Areas adjacent to highways that are more likely to be used despite the presence of the roadway may also be the locations where animals are most likely to cross the road.
The quality of roadside habitat in relationship to a species preferred habitat type is one factor that appears to play a role in an animals willingness to approach roads. Grizzly bears in Alberta, Canada (Gibeau et al., 2002) and bobcats in Wisconsin (Lovallo and Anderson 1996) were both more likely to use areas near roads when these roads were located in preferred habitat types. The response of grizzly bears to preferred habitats was weaker along roads with high traffic volumes (Gibeau 2002). Lovallo and Anderson (1996) did not describe the characteristics of roads in their study, nor did they separate them by type in their analysis. Food and the presence of cover in the form of trees are also two habitat factors that influence roadside habitat use. Mule deer in Colorado avoided all roads, including highways, under all conditions, but showed a reduced avoidance response in pine forests and juniper woodlands, as compared to shrub habitats (Rost and Bailey 1979). Elk in Montana showed a similar pattern, increasing their use of roadside habitats as canopy cover increased (Lyon 1979). Roadside habitat use by white-tailed deer (Odocoileus
30


virginianus) in Pennsylvania was most influenced by the presence of food. In locations where the roadside habitat was primarily wooded, deer often entered the right-of-way to graze on grasses growing along the shoulder and in the median. In locations where the roadside habitat was primarily open, deer foraged away from the right-of-way (Carbaugh et al., 1975, Feldhamer et al., 1986).
Animal/Vchicle Collision Locations Alone Highways Although accounts detailing the phenomenon of roadkill in the U.S. were published almost as soon as widespread use of the automobile began (e.g., Stoner 1925, Cottam 1931), researchers did not undertake formal studies about AVCs until the early 1970s. These initial studies focused on white-tailed deer and were conducted in Pennsylvania (Beilis and Graves 1971, Puglisi et al., 1974) and Michigan (Reilly and Green 1974, Allen and McCullough, 1976.) Because the focus of these studies was on the temporal rather than the spatial distribution of collisions, the information they contain regarding AVC locations is qualitative rather then quantitative. However, the authors of all four studies concur that high kill areas were associated with locations that had cover types that attracted foraging deer, as well as high traffic volumes.
Investigations conducted since the 1980 have utilized more rigorous approaches to quantify the features associated with AVC locations. Using logistic regression, Bashore et al. (1985) determined that variables representing longer driver line-of-sight distances and a mix of cover types played the primary role in identifying higher density deer roadkill zones on Pennsylvania highways. They hypothesized that long sight-lines encourage motorists to drive faster, making it difficult for them to avoid deer on the roadway, and that mixed cover types attract more foraging deer because they provided both food and cover.
31


The most recent studies used GIS to analyze the characteristics of AVC locations. Finder et al. (1999) used remotely sensed site and landscape data to determine that forest cover was the variable most strongly associated with high kill zones in Illinois. In addition, gullies adjacent to the road, riparian zones, and public recreational land within 0.8 km of a point further increased the probability that it would be classified as high kill. Hubbard et al. (2000) analyzed deer/vehicle collision locations in Iowa using a similar approach, and determined that collisions increased as cover type patch size, number of bridges per 1.61 km highway segment, and lanes of traffic increased.
Although the locations of AVC provide an indication of where wild animals interact with highways, they may be a poor surrogate for determining where animals prefer to cross highways. Animal/vehicle collision data record only unsuccessful crossings attempts, and only a small subset of those. AVCs are generally reported only when property damage or injury to the vehicles occupants warrants a call to the police. Based on estimates from state transportation and wildlife management agency field personnel, Romin and Bissonette (1996a) reported that nationwide, only 17 to 50 percent of AVCs that occur are reported. Additionally, because responding officers usually estimate the location of the AVCs to the nearest milepost, only the general location of wildlife highwaycrossing zones can be gleaned from this data.
Thus, the locations of actual road-killed wildlife observed in the field, instead of reported AVC, may be a more accurate indicator of the locations where wildlife interact with highways. Two published studies address the factors associated with lethal interactions, as indicated by carcasses. The distribution of road-killed mule deer in Utah was positively correlated to locations where large drainages bisected the road and to areas with non-agricultural cover types (Romin and Bissonette
32


1996b). Inbar and Mayer (1999) used multiple regression and correlation analyses to determine that high mortality areas for armadillos (Dosypus novemcinctus) in Florida were associated with preferred habitat types in winter and areas with high night-time traffic volumes in summer.
Use of Highway-Crossing Structures by Wildlife
Wild animals that cross highways can do so either at-grade, by walking across the hard top, or they may utilize structures to cross over or under die roads surface. The bulk of the existing research pertaining to the locations where wildlife cross highways is focused on animals use of such structures. This research looks at use of structures constructed specifically for wildlife as well as at opportunistic use of structures installed for other reasons. Studies have been conducted along both fenced (Foster and Humphrey1996, Roof and Woodling 1996, Clevenger and Waltho 2000) and unfenced highways (Hunt et al., 1987, Yanes et al., 1995, Rodriguez et al., 1996a, Hewitt et al., 1998, Haas 2000, Clevenger et al., 2001). In aggregate, these studies indicate that animals use a wide variety of structures, including bridges, culverts, and tunnels, as well as the structures built specifically for wildlife, to cross under highways and high-speed rail corridors.
However, studies focused on underpass use, along either fenced or unfenced highways, cannot indicate anything about where animals prefer to cross transportation corridors at-grade. Studies conducted along highways fenced to prevent crossing except at structures can only test the preference for different structures and structure location. Wildlife using structures along unfenced highways and railways can cross at any location, but in all the existing studies, data were collected only at the structures. Therefore, conclusions can still only be made about structure preference.
33


Research about overpass use by wildlife is very limited and suffers from the same limitation as the underpass research. Wildlife-friendly overpasses have been built in only two locations in North America (Banff National Park, Alberta, Canada, and I-15 in Utah, USA) and research regarding use of these structures is unavailable. Overpasses have been more widely constructed in Europe, but the single available English-language report summarizing their use indicates that only the structures were monitored (Pfister 1997). Thus, like the underpass research, no conclusions can be drawn from it about where animals prefer to cross the highways in question at-grade.
Never-the-less, the results of these studies do provide some insight into the habitat variables associated with structures most frequently used, and it is possible to infer that some of these same variables may influence the choice of at-grade crossing locations as well. Results of research that quantitatively compared crossing rates and the surrounding habitat at multiple structures indicate that habitat elements from both the local- and the landscape-scale play a role in determining a structures rate of use.
At the landscape-scale, Haas (2000) found that species sensitive to human- induced habitat fragmentation use underpasses associated with such habitats less than the underpasses associated with more intact habitats. At the local-scale, researchers correlated a variety of habitat features to underpass use. The level of human activity is a significant factor for some species. Predators, but not ungulates, tend to avoid structures with high levels of human activity (Rodriguez et al., 1996a, Clevenger and Waltho 2000).
The presence of cover at or near the entrance to an underpass increases the rate of use by many small- and mid-sized species, including mustelids, canids, felids and
34


lagomorphs (Yanes et al., 1995, Rodriguez et al., 1996a, Haas 2000, Clevenger et al., 2001). The characteristics of the roadway itself, such as pavement width, traffic volume, and associated noise levels are also important to these small- and midsized species (Yanes et al., 1996, Clevenger et al., 2001). Response to these variables varied by species and the authors hypothesized that this was due to each species vulnerability to predation while crossing the clear zone associated with the roadway. These local-scale habitat factors appear to override structural characteristics of the crossing structure itself in some cases, especially for predator species (Rodriguez et al., 1996a, Yanes et al., 1996, Clevenger and Waltho 2000).
Locations of At-Grade Highway Crossings bv Wildlife
Studying at-grade animal movements along an entire highway corridor, as opposed to concentrating on focal points such as underpasses, is difficult. Methods for recording corridor-wide data include direct observation of animals crossing the highway, continuous monitoring of radio-collared animals, and using tracking techniques to infer animal movements.
Direct observation is not a very efficient approach, as it is labor intensive and subject to bias animals will only be seen crossing roads in location that are under observation during that particular event. However, while conducting a study of roadside habitat use by white-tailed deer, Carbaugh et al. (1975) opportunistically observed 160 deer in the act of crossing the eastbound lanes of 1-80 in central Pennsylvania. The authors did not provide a detailed description of the highway (lane width, median width, etc.), and did not analyze the crossing data quantitatively. Nevertheless, they provide a valuable qualitative description of crossing behavior. Observed crossings were focused on areas that had not been modified during highway construction by cutting and filling. Hill-valley intergrades
35


and areas that sloped upward from the highway were used most often. The authors hypothesized that deer were guided to these unmodified areas by cuts, fills and guardrails. Deer rarely jumped guardrails to enter the roadway, although they jumped them readily to exit it.
Continuous monitoring of radio-collared animals provides another option for recording highway their crossing locations. This approach is also labor-intensive when using standard VHF collars. Consequently, the staffing constraints of most wildlife studies would require choosing a single focal animal to follow during each monitoring session, resulting in bias and limited data sets. Therefore, this method apparently has not been widely applied. A study of highway crossing behavior by wolves in Wisconsin (Kohn et al., 1997) that used this technique did not get an adequate sample size of crossing events for analysis. The advent of radio collars that can continuously relay a GPS derived location to a receiver via satellite is making continuous monitoring of an animals movements more practical. However, the application of this technique is currently limited, as these transmitting devises require significant battery power. Because large batteries are heavy, these types of collars can only be placed on large animals, such as bears and bison. A study of grizzly bear movements in relationship to roads in Montana is in progress (Waller and Servheen 1999), but results are not yet available.
Roadside tracking techniques provide a more versatile option for recording animal movements across roads and provide a reasonable trade-off between labor required and data acquired. However, a naturally present, suitable roadside tracking medium is needed throughout the corridor under study to record animal movements, because creating a medium in such large areas is infeasible. Reliable snow cover during the winter months provides perhaps the best opportunity to conduct this type of research, and three studies, (Mamalis 1995, Singleton and Lehmkuhl 2000,
36


Alexander and Waters 2000) have taken advantage of this opportunity. Mamalis (1995) recorded crossing behavior of all species of wildlife present along the Trans-Canada Highway in Banff National Park, Alberta, during the January-March period of 1995. The distribution of crossings was significantly different from random (x2 =15.51, 8 d.f., p<0.05) and crossings were more common than expected in several locations. Snow depth, the only factor that was quantitatively assessed, did not significantly influence choice of crossing location. Qualitative descriptions of crossing zones indicated that existing wildlife travel corridors and roadside barriers might have influenced where animals crossed the highway.
Singleton and Lehmkuhl (2000) recorded crossing location of all species present from January through march 1999 and from December 1999 through March 2000 along a 30 mile stretch of 1-90 between Cle Elum and Snoqualmie Pass in Washington, USA. They mapped their results and visually identified distinct clusters of crossings, but did not quantitatively analyze the extent of the clusters or the habitat features associated with them. Qualitative descriptions of these locations indicate that they were generally associated with intact natural cover types, low levels of human disturbance, and landscape linkage zones that the authors had previously identified through a GIS-based analysis.
Alexander and Waters (2000) recorded crossing behavior of all species of wildlife present along three 30 km sections of highway in Banff National Park, Alberta, Canada during the winters of 1997/1998 and 1998/1999. They identified high frequency crossing zones ranging from 250 to 2000 m in diameter. The authors then used chi-square tests to compare the characteristics of crossing locations to expected values. Results indicated that general predictors of movement for all species included slopes facing south, southwest and west, slopes angles of 5 or less, and areas of low topographic complexity. However, the authors did not
37


indicate at what scale these measurements were made, i.e., they did not state how far away from the roads edge their measurements of the surrounding habitat extended.
Summary
Reviews of literature pertaining directly to where wild animals choose to cross roads indicate that at-grade crossings do not occur at random locations. This conclusion is further supported by the non-random distribution of AVCs and roadkilled carcasses, both of which can be viewed as indicators of crossing behavior, and by the preferential use of crossing structures associated with certain habitat features. Additionally, non-random variation in the intensity of roadside habitat use also suggests that wild animals are more likely to approach highways in certain locations.
The literature pertaining to each of these topics indicates that a variety of factors Influence how and where wild animals interact with the roadside. The most important variables identified in this literature review were the presence of food, cover, preferred habitat, drainages that intersect the roadway, and characteristics of the roadway itself. Animals also avoided roads in areas with high levels of habitat fragmentation and, for many species, high levels of human activity. Although only one of these factors, habitat fragmentation, was explicitly investigated in terms of scale, it is also clear that variables from both the local and the landscape scale play a role in the location of wildlife/highway interactions. However, because of the disparate study designs used in the reviewed literature, it is not possible to compare the influence of local- versus landscape-scale variables, or to rank these variables in order of importance.
38


CHAPTER 3
IDENTIFYING WILDLIFE HIGHWAY-CROSSING ZONES IN THE COLORADO ROCKIES: RESEARCH BACKGROUND
Introduction
Overview
The goal of my research was two-fold. First, I wanted to determine if the locations where wild animals cross highways are correlated, with definable characteristics of the surrounding environment and the roadway itself. Then, I was interested in how this type of information could inform the highway design process to reduce wildlife/highway conflicts. I hypothesized that:
Wild animals do not cross highways at random and the locations that they use can be correlated with features of the roadside and the surrounding habitat.
The identification of likely crossing locations is an accessible process that can be incorporated into highway design to reduce wildlife/highway conflicts.
In this chapter I describe the rationale for and methodology of the research that I conducted to examine the first part of my hypothesis. I begin by reviewing my approach, then justify the comparison variables that I selected, and describe my study sites. The remainder of the chapter is dedicated to describing my methods of data collection and analysis in detail. The data collection descriptions are arranged
39


according to the type of data that was collected. Explanations of the data analysis are arranged according to the question being examined. The arrangement of the analysis descriptions serves also to organize Chapter 4 (results) and Chapter 5 (discussion of the results). The second part of my hypothesis is examined in Chapter 6.
The Research Approach
My research identified locations where animals crossed unfenced highways at-grade. I then compared the crossing locations to random locations along the highway, to determine if and how these two types of locations differed. I defined highways as paved roads, at least two lanes in width, and with posted speeds of 55 km/h or greater. Other types of roads (e.g., unpaved, single lane, and/or low speed) have a variety of effects on wildlife, including behavior modification, introduction of competitors, and increased human activity (Trombulak and Frissell 2000), but are apparently readily crossed by wildlife (C. Apps, pers com 1998; McKelvey et al., 1999). Therefore, I did not consider these types of roads.
Features that may differ between crossing zones (CZs) and random locations could come from either the local or the landscapescale. Local and landscape are relative terms that must be defined by the context of their application. For this study, I considered the local scale to be the characteristics of the roadway itself as well as topographic and vegetation features within 100 m of the roadway. I defined the landscape scale as the general landforms and cover types encompassed by the ridgelines that provided visual boundaries of the area surrounding the highways in my study areas.
40


To identify CZs for comparison with random roadside locations, I recorded animal activity at the roadside along two highway corridors, one with a two-lane highway and one with a four-lane highway. I chose two highways with different footprints because I wanted to know if highway design affected the highway-crossing locations of wildlife. I analyzed animal roadside activity data to identify the areas that wild animals crossing the roadway used most often. I monitored roadside wildlife activity along the entire length of both highway corridors year-round, as opposed to seasonally and/or only at focal points such as underpasses. Additionally, I monitored bridges and over-sized culverts to determine if animals were actually using them cross under the highway, and whether the presence of an underpass inflenced at-grade crossing rates and locations.
Relevance of the Measured Variables
I chose the variables to compare crossing with random locations by considering the existing research pertaining to wildlife/highway interactions. In addition, I considered the suite of roadside habitat characteristics whose form, and presence or absence of which are under the direct control of highway designers and builders. A complete list of all the variables that I measured is given in Table 3.1, and the rationale for their inclusion is discussed below.
Although I chose to categorize variables as landscape-scale or local-scale (Table 3.1) when describing study methods and results, it is important to note that they could be divided in other ways. Other classification schemes include habitat characteristics versus highway characteristics and measurements that must be made in the field versus measurements that can be made remotely from digital data layers or aerial photographs. Habitat characteristics are a mix of local- and landscape-scale variables, where as highway characteristics are, by the definition of
41


my study, local scale only. The measurements that I made in the field for this study were also all at a local-scale, but the remote measurements encompassed both local- and all landscape-scale measurements.
Table 3.1 Variables measured to compare crossing-zone (CZ) locations to random locations
Variable Description Local or Landscape Scale Habitat or Highway Characteristic Field or Remote*
Distribution of cover types throughout study site Landscape Habitat Remote
Distribution of slope classes throughout study site Landscape Habitat Remote
Distribution of aspect classes throughout study site Average size of cover type patch Landscape Habitat Remote
Landscape Habitat Remote
Average size of slope class patch Landscape Habitat Remote
Average size of aspect class patch Landscape Habitat Remote
Average complexity of cover type patch Landscape Habitat Remote
Average complexity of slope class patch Landscape Habitat Remote
Average complexity of aspect class patch Landscape Habitat Remote
Line-of-Site (m) along roadway Local Highway Field
Distance (m) to nearest woody vegetation at the Local Habitat Field
roadside
Line-of-sight to roadway 20 m from the roadside Local Habitat Field
Distance to nearest woody vegetation 20 m from Local Habitat Field
the roadside
Minimum and maximum slope at roadside Local Habitat Field
Distance to nearest center of human activity Local Habitat Field
Distance to nearest side road Local Habitat Field
Distance to nearest drainage Local Habitat Field
Distance to nearest roadside barrier Local Highway Field
Distance to the forest edge Local Habitat Remote
Location of CZ relative to underpass Local Highway Field
Cover types associated with CZs compare to Local Habitat Remote
available cover types
Slope classes associated with CZs compare to Local Habitat Remote
available slope classes
Aspect classes associated with CZs compare to Local Habitat Remote
Available aspect classes
Measured in the field or from remote photography
Landscape Scale Variables
Existing research pertaining to wildlife/highway interactions, reviewed in Chapter 2, suggests three main landscape-scale habitat characteristics that influence how wild animals move through the landscape, and consequently, where they approach the roadside. These characteristics include topography (e.g., Romin and Bissonette
42


1996; Finder et al., 1999, Alexander and Waters 2000, Hubbard et al., 2000); avoidance of human activity (e.g., Rodriguez et al., 1996a; Clevenger and Waltho 2000); and the continuity and heterogeneity of cover types (Haas 2000; Hubbard et al., 2000).
Topography, as defined by the studies cited above, encompasses the general slope and aspect of a landscape as well as the landforms it contains, such as drainages and ridgelines. Slope and aspect are important because most species avoid excessively steep slopes, and some species may have a preference for a certain aspect. Additionally, topographically complex landscapes (i.e., those with many change in slope and aspect over a small area) may also be difficult to travel through. According to the studies cited above, drainages influence how animals move through the landscape and consequently where they come to the roadside. It is unclear if that is because they act as guideways due to their linear form or because animals prefer the habitats available along them.
Landscape-scale continuity and heterogeneity of cover type may also influence where animals will approach the roadside, but their role is unclear. Hubbard et al. (2000) found that deer were more likely to approach the roadside in areas with large, contiguous blocks of cover. However, in a simple model of animal movement through habitat patch mosaics, Stamps et al. (1987) showed that the rate of movement should be highest in mosaics comprised of many patches with high edge-to-size ratios, as long as the edge is moderately soft. Edge softness is defined by a species ability to move between the habitats that the edge separate. Hubbard et al. (2000) also suggested that distinct edges between different cover types may act as linear guideways, much like drainages and ridgelines. This intriguing idea does not appear to have been addressed in the habitat connectivity
43


literature, nor has it been a topic of edge-effect research (reviewed in Lidicker, 1999).
Based on the literature cited above, I chose two types of landscape-scale variables. First, I measured the composition of the landscape, based on the amount of each cover, slope, and aspect class surrounding the highways. Then I measured complexity of the landscape. Metrics of landscape structure that can be used to assess complexity include patch size and patch edge-to-size ratio (ESR). A landscape made up of many small patches is more complex than is a landscape made up of a few large patches. Additionally, a group of patches with small average sizes and long edges relative to their size represent a more complex landscape structure then a group of patches with large areas and shorter relative edge lengths.
I did not quantitatively assess the effect of human activity or linear guideways such as drainages, at the landscape-scale. I purposefully chose study sites surrounded by public land. Therefore, most of the human activity that occurred at these sites was limited to the roadside and I considered avoidance of human activity a local-scale variable. I limit my examination of linear guideways to a qualitative assessment because I had no objective methodology for identifying what constituted a linear feature.
Local-Scale Variables
Existing research pertaining to wildlife/highway interactions, reviewed in Chapter 2, suggests a suite of local-scale habitat characteristics that influence how animals use roadsides. These variables include the presence and location of roadside barriers, food, and preferred habitat (see Table 2.1). Avoidance of human activity
44


and topography, specifically drainages that intersect the roadway, are also thought to influence how animals use roadsides (see Table 2.1). The form, presence, and absence of some of these roadway and roadside habitat characteristics are under the direct control of highway designers and builders. These variables include the amount of cover left in or planted in the right-of-way (ROW), the distance from the roadside to cover, placement of roadside barriers, and the location and configuration of crossing structures, such as underpasses.
Based on the information discussed above, I chose local-scale variables that described the relationship of topography, cover, roadside barriers, underpasses, and human activity to locations that animals used to cross roads. I excluded measurements related to the presence of food and preferred habitat because I wanted to focus on variables that all medium- to large-sized species were likely to perceive, and these variables are species specific.
Study Site Descriptions
I selected two study sites in the Southern Rocky Mountains of Colorado, USA. One was located along US 24 at Trout Creek Pass and the other along 1-70 at Vail Pass. My primary study site selection criteria included: reasonable proximity to Denver, CO, my home base; the presence of a suitable roadside tracking medium year-round; and a wide shoulder to provide a safe work environment. Another consideration was adjacent public land, to reduce potentially confounding effects of human disturbance associated with homes and businesses and to help ensure a reasonably large local wildlife population.
45


Trout Creek Pass
The Trout Creek Pass study site (TCP) was located predominantly in Chaffee County, CO, USA and encompassed 11.0 miles (17.8 km) of narrow, two-lane highway that simultaneously serves as US 24 and US 285. The mile postings (MP) in this stretch of road follow the sequence of US 24 and range from MP 216.0 approximately 2 miles east of Johnson Village, to MP 226.0, approximately one mile east of Trout Creek Pass (Figures 3.1, 3.2). A small section of the study area, to the east of the Pass, was located in Park County. US 24 is a two-lane road throughout the study area, except for the east side of the Pass where a climbing lane creates a short section with three lanes. Lanes are 3.7 m wide and shoulders are unpaved. The average annual daily traffic volume is 4000 vehicles (CDOT 2000).
Although US 24 is classified as an east-west highway, in the north end of TCP it runs predominately north-south (MP 221-226). The terrain in this part of the study area was rolling, and cover type consisted of open grasslands and shrub communities west of US 24, and mixed coniferous forests to the east. The shrubs at TCP were tall, and like the forest cover types, they provided good cover. The terrain in the south end of the study area was rugged and highly dissected by dry washes and rocky outcrops, and the vegetation consisted of coniferous stands intermixed with aspen, deciduous shrubs, and small, open grassy areas. Additionally, a narrow shrubby riparian zone associated with Trout Creek, which paralleled the highway in the southern section of the study area. Elevations in the study area range from 2830 m at the Pass to 2420 m at MP 216, and the main source of human disturbance, apart from the highway itself, were about 20 homes located mainly in the southern end of the study area.
46


Figure 3.1 The location of the Trout Creek Pass study site (TCP) in Chaffee County, Colorado.
47


Figure 3.2 Detail of the Trout Creek Pass area, location of TCP.
48


US 24 intersected six major drainages in the study area, which were bridged by large, three-chambered concrete box culverts with concrete floors or by smaller bridge structures with natural floors. Numerous smaller drainages in the study area were bridged with pipes or culverts of various sizes, ranging from 0.03 to 1.0 m in diameter. Because snow cover in the study area was usually temporary, this area acted as both summer and winter range for mule deer and elk, although elk were more common at the north end and in winter. Other common terrestrial species included red fox (Vulpes wipes), gray fox (Urocyon cinereoargenteus), coyote (Cams latrans), mountain lion, bobcat, long-tailed and short-tailed weasel (Mustela firenata, M. erminea), and mountain cottontail (Sylvilagus nuttallii).
The tracking medium at TCP consisted mainly of native roadside dirt, sand, and gravel, which provided an inconsistent tracking surface throughout the study area. Some locations had very fine-grained dirt along the shoulder and readily took impressions from both hoofed and pad-footed animals. Most locations however, had a mix of dirt, sand and gravel and recorded hooves far more efficiently. The variability in the tracking medium was exacerbated by variations in its ability to retain moisture. Areas with more gravel dried more quickly and yielded poorer quality track impressions. In most places, the unvegetated area along the roadside was about a meter wide.
Vail Pass
Straddling Summit and Eagle counties, CO, USA, the Vail Pass study site encompassed 12.0 miles (19.4 km) of 1-70, from MP 183.0 to MP 195.0 (Figures 3.3, 3.4). Because of the heavy winter snows this site receives, the suite of animal species present (described below) and their behavior differs substantially between the times of the year when snow is present compared with when snow is not
49


Figure 3.3 Location of the Vail Pass (VP) and Vail Pass Snow (VPS) study sites, straddling Eagle and Summit Counties in Colorado.
50


Figure 3.4 Detail of the Vail Pass area, location of both VP and VPS, including the location of Copper Mountain Resort (CMR).
51


present. Therefore, I considered the Vail Pass area to be two different sites, based on snow depths. I will refer to the site as Vail Pass (VP) when the ground was snow-free and as Vail Pass Snow (VPS) when there was snow on the ground.
The highway at VP and VPS is designated as an east-west road and the average annual daily traffic volume is 15,500 vehicles (CDOT 2000). Vail Pass, located at approximately MP 190, divides the study area into an east side and a west side. The section of 1-70 encompassed by the study site contained two 4.1 m-wide lanes and 4.7 m of associated paved shoulders for a total of width of 12.9 m in each direction. The alignments of the east and westbound lanes were independently sited and varied in location and elevation. The median separating the east and westbound lanes varied in width from less then a meter in some places on the west side of the Pass, up to 260 m on the east side of the Pass. To a large extent, the natural cover and topography were maintained within the wide median area on the east side. On the west side, Jersey barriers separated the east and westbound lanes in locations where they were at the same elevation. Stepped retaining walls were used to separate the lanes in locations where one lane was at a higher elevation then the other. On both the east and west sides, additional Jersey barriers were used along the outer shoulder in locations where steep drop-offs occurred. In general, there were far fewer vertical roadside and median barriers on the east side than on the west side of the study site (Figure 3.5).
1-70 intersected 18 large drainages in the study area, and bridges spanned 11 of them. These bridges provided high quality highway crossing opportunities for wildlife as the drainages they spanned are wide (up to 230 m), and the natural cover below most was largely undisturbed. Most of the smaller drainages that intersected the roadway in the Vail Pass area were diverted into pipes less then 1 m in
52


Figure 3.5 The locations of all barriers and underpasses on 1-70 at Vail Pass. Note the distribution of barriers and the alignment of the underpasses on the two sides of the Pass.
53


diameter. The primary cover type in the Vail Pass area was mixed coniferous forest interspersed with aspen stands, sub-alpine meadows, and willow carrs. These shrubby willow areas did not provide good cover. The elevation of the study site ranged from 2730 to 3165 m, and sources of human-induced disturbance, aside from the highway itself, included a rest area, truck turn out and maintenance shed at the summit and the Copper Mountain Resort at the base of the east side. Common terrestrial wildlife species in this study area included red fox, bobcat, mule deer, elk, and mountain lion during the snow-free months. Snowshoe hare (Lepus americanus), coyote, long-tailed and short-tailed weasels, and American marten {Maries americana) were present year-round.
The tracking medium at VP consisted mainly of road sand left over from the previous winter. It provided a uniform tracking surface throughout the study area. The sand was a mix of fine and coarse grains, and took the imprint of hooves readily, but was less efficient at recording the passage of pad-footed animals. Frequent thunderstorms kept the roadside at VP relatively moist throughout the summer, further improving its quality as a tracking medium. The thick layer of road sand also choked out most roadside plants, and an unvegetated swath of sand generally extended from l-3m away from the paved shoulder. At VPS, I used snow as the tracking medium.
Data Collection Methods: Tracking
Standard Tracking Methods
I recorded locations throughout all three study areas where medium- and largesized mammals (mule deer, elk, coyote, fox, bobcat, mountain lion) crossed the highway, as indicated by their tracks. At TCP and VP I checked roadside transects
54


200 m in length for tracks during each field session. To ensure that transects were distributed throughout a study area and did not overlap, I used a stratified random selection approach, varying transect location for each data collection session. At each transect, a field assistant or I walked along the highway at the pavements edge and looked for tracks left in the unpaved shoulder. At TCP, traffic was light and I crossed the highway to walk along both sides of it, and recorded tracks from both sides. At VP however, I only walked along the outer edges of the west- and east-bound lanes. Due to high traffic volumes and speeds, I considered crossing the highway to access the median-side roadside unsafe.
I recorded track locations using a hand-held GPS device\data logger (Geo Explorer II, Trimble) that automatically recorded location while I entered information through a menu-driven interface. All tracks of the same species observed within a 5-meter stretch were recorded as a single track record (TR). Each TR contained the following information: species of animal, number of animals, location (UTM coordinates), activity (described below), surface (dirt or snow) and date. I downloaded data files from the data logger and used Trimbles proprietary software to convert them to Excel spreadsheet and ArcView shapefile formats for analysis.
I interpreted the activity of an animal from the pattern of tracks it left behind. Activity included four classifications: 1) Crossing a track pattern indicating the animal passed across the roadway from one side to the other; 2) Approach a track pattern indicating the animal approached the roadway but did not pass across it to the other side; 3) Parallel a track pattern indicating the animal walked along the roadside for a distance of 10 m or more; and 4) Undefined a track pattern that did not clearly indicate any of the three behaviors described above. At TCP, I only classified a TR as crossing if a matched set of tracks was found on both sides of the road. At VP, I did not confirm crossings in this way because I considered it
55


unsafe to cross the roadway. Instead, I classified track sets that were perpendicular to the road and did not have a matched set within 20 m going in the opposite direction as crossing. Because the swath of tracking medium along the roadside was wide at VP, it was usually possible to read an animals behavior at the roadside quite clearly and I only designated a TR as crossing when I was reasonably sure the animal had indeed passed to the other side of the roadway.
At TCP, I collected data twice weekly from September through June, weather permitting. I collected track data only once a week during July and August because animals moved out of the area during summer, and very few tracks were found during this period. Data were collected along 11 randomly chosen 200 m transects and existing roadside substrates (fine-grained dirt, sand, or snow) used as a tracking medium. At VP, I collected data using the protocols described above from June through November. Data were collected twice weekly, weather permitting, along 10 randomly chosen 200 m transects and existing roadside substrates (fine-grained dirt or sand) used as a tracking medium.
Standard Underpass Monitoring
In addition to monitoring the roadside for tracks, I monitored some highway structures (bridges, oversized concrete box-culverts) at both study sites that could have been used by animals to cross under the highway. Although only one of the monitored structures was constructed specifically to act as a highway underpass for wildlife, I will refer to all these structures as underpasses. All underpasses monitored spanned either narrow perennial streams or intermittent drainages that only carried water during spring run-off or during storm events, and offered plenty of dry substrate for animals to use when they passed through. I created track beds from locally available sand and soil at both ends of each monitored structure. An
56


animal was recorded as passing through a structure only when I observed a matched set of tracks at both ends.
At TCP, a field assistant or I monitored a subset of 10 underpasses, chosen based on accessibility and safety considerations. Due to time constraints only two underpasses were randomly chosen and checked during each data collection session. Track beds and the roadside within 100 m of either side of the structure were checked for tracks to determine if animals crossing at that location had crossed at-grade or used the structure. Additionally, I checked the track beds, but not the adjacent roadside, in the structures located in the drainages at MP 215.0, MP 216.5, Shields Gulch (MP 218.4), and Magee Gulch (MP219.2) as often as time permitted. I checked these four large drainages often because I assumed they were most likely to be used, and I wanted to record the full variety of animals willing to cross under the unfenced road at TCP. At VP, I monitored four of the 17 underpasses; the other 13 were deemed unsuitable for monitoring due to safety considerations, high levels of human use, or excessively large size, which made maintaining the trackbed difficult. I checked track beds and the roadside within 50 m of either side of each underpass for tracks to record if animals crossing at that location had crossed at-grade or used the structure. All four underpasses were checked for tracks as a part of every data collection session.
Snow Tracking Methods
Snow is a superior tracking medium. Snow cover allows tracks to be spotted much more easily and allows the tracks of a wider range of species to be observed. In addition to the tracks of large species that were observed when using roadside dirt/sand as tracking medium, I recorded smaller species such as snowshoe hare, cottontail rabbit, fox, weasel, and marten from snow. Snow tracking was conducted
57


at VPS December through March during both 2000/2001 and 2001/2001.1 did not implement snow-tracking protocols at TCP because sufficient snow cover at this site was infrequent, unpredictable, and ephemeral. Thus, even on the few occasions when there was snow on the ground at TCP, the standard tracking procedures described above were followed. However, the snow did present the opportunity to record fox and rabbit tracks at TCP, which rarely left readable tracks at the roadside otherwise.
Using snow-tracking methods, I observed the entire VPS study area, as opposed to a subset of transects, for tracks. Due to the snow depths, far fewer animals are present in the Vail Pass area during winter than during the summer. Thus, finding and recording all trails present was a reasonable task. A field assistant or I located all animal trails that entered the roadway within the study area by driving slowly (< 25 km/h) along the shoulder. When a trail was observed it was identified by species, and crossing success determined. Using the GPS device/data logger, I recorded all information, including species, activity, number of individuals, substrate, and the UTM coordinates of the trails intersection with the highway. Additionally, I monitored five underpasses for animal use during the winter, as weather permitted. The trails of all animals entering the space underneath these structures were observed in the snow, and each animals behavior was recorded as either passing through or not passing through.
The number and timing of snow tracking sessions was weather-dependent. I only conducted snow tracking after a recent snowfall. Storms occur roughly every four to eight days over Vail Pass during the winter. High winds, warm temperatures, and sunny days, all of which can render tracks unreadable, are common on Vail Pass throughout the winter, so it was necessary to track as soon as possible after a storm. I conducted most roadside tracking sessions 6-32 hours after a snowfall. The terrain
58


under an underpass and the underpass itself protects tracks from wind and sun. Therefore, I waited to check underpasses for tracks until after roadside tracking was completed, usually 32-58 hours post snowfall. Depending on the number of tracks located, one to two days were required to conduct the roadside tracking, and structure monitoring required an additional day of work. I did not conduct snow tracking and structure monitoring while snow was falling. Occasionally there was insufficient time between storms for a complete 2- to 3-day data collection cycle. Additionally, due to heavy traffic volumes, roadside work was deemed unsafe on weekends and holidays, and only structure monitoring was conducted on these days.
Animal Abundance
To determine how local animal densities might influence crossing rates, I attempted to estimate the relative abundance of animals in the immediate vicinity of the roadway at both study sites. Within 36 to 60 hours of a snowstorm, I walked transects 1.6 km long parallel to and 300 m from the roadside, and counted all animal trails that intersected my path. Because other snow tracking tasks also had to be preformed in this timeframe, it was difficult to schedule these off-road transects. Additionally, these transects usually required breaking trail in deep snow over rough terrain. Therefore, only one or two could be walked in a single day. Because of these constraints, only five off-road transects were walked at VPS and 14 at TCP.
In addition to collecting field data, I also contacted the Colorado Division of Wildlife (CDOW) and obtained statewide mule deer and elk population estimates. The CDOWs data reports population estimates by Game Management Units (GMUs), areas that the DOW believes are relatively homogenous in both the
59


habitat quality and the population density of the species of interest. The boundaries of these GMUs differ for mule deer and elk. Although the GMUs are many times larger than my study sites, these population estimates provided a useful indication of the relative densities of the mule deer and elk between the two sites.
Data Collection Methods: Habitat Measurements
Landscape-Scale Habitat Measurements
I made all measurements of landscape structure for landscape-scale comparisons from digital data layers, using the ArcView software package. These data layers were generated from aerial and satellite photography, and I used coverages created by the U.S. Geological Survey (USGS) when available, including the National Land Cover Data (NLCD) for TCP, and 10-meter contour resolution Digital Elevation Models (DEMs) at VP and VPS. Because my field experience suggested that the NLCD coverage of the Vail Pass area misclassified the cover type of large areas, I used a digital vegetation map created by the Forest Service for vegetation measurements at VP and VPS. For the TCP area, 10 m resolution DEMs were not available from the USGS. I commissioned the Remote Sensing and Geographic Information Group (RSGIG) located in the Denver, CO, office of the U.S. Bureau of Reclamation to create a 10 m DEM of the TCP site specifically for my project. This DEM was created by combining survey control points of known elevations with existing stereo photography using the ERDAS Orthobase software package.
At all three study sites, the landscape I measured was the area encompassed by the ridgelines that provided visual boundaries surrounding the highway. I derived the vegetation patterns of these landscapes from the digital cover maps and the topographic patterns from the DEMs, The NLCD is divided into 21 cover type
60


classes in Colorado, and I reclassified the Forest Service data layer to match those classes. I used the ArcView extension Spatial Analyst to process the DEMs into 19 slope classes and nine aspect classes. I classified topographically level areas as flat, then divided slope by 5-degree increments for 18 additional classes and aspect by increments of 45 degrees for eight additional classes (Table 3.2).
Table 3.2 Definitions used to divide cover, slope, and aspect classes on digital maps of the three study sites
Digital Coverage Type Class Definitions
Cover Type Water Ice and Snow
Commercial/Transportation Low Density Residential
Barren (rock or sand) High Density Residential
Quanies Transitional
Deciduous Forest Orchard
Coniferous Forest Hay
Mixed Forest Row Crops
Shrubland Small Grains
Grass] and/Herbaceous Fallow
Emergent Wetlands Woody Wetlands Urban Recreational
Slope Flat 0-5 46-50 6-10 51-55 11-15 56-60 16-20 61-65 21-25 66-70 26-30 71-75 31-35 76-80 36-40 81-85 41-45 86-90
Aspect Flat Southeast
North South
Northeast Southwest
East West Northwest
I was careful to choose digital base maps that adequately reflected actual land cover classes and their boundaries. I based my assessment on my familiarity with the study sites. I wanted to be confident that the maps reflected variations animals
61


could perceive and respond to. The NLCD is divided into 21 classes in the Colorado, 10 of which are naturally occurring land cover types (Table 3.2).
Because these classes are broadly defined, I believe medium- to large-sized mammals can readily perceive the variation they represent, and this classification scheme was reasonable for my analysis. Similarly, I chose the slope and aspect increments (Table 3.2) because my experience with animal behavior and habitat selection suggests that animals respond to topographic variation at those levels.
Local-Scale Habitat Measurements
I collected two types of local-scale habitat data in the field from both study site locations. The first type of local-scale data that I collected was the characteristics of point locations within crossing zone (CZs) as well as the characteristics of random points for comparison. The methods I used to identify CZs are described in the Data Analysis section of this chapter. I chose point locations within the CZs by visually inspecting maps of the CZs, and locating points where the highest density of crossing TRs occurred. I chose random point locations with ArcView, using a script that generated random points along lines representing the highways in the study areas. Using ArcView, I determined the UTM coordinates of both random and crossing points and identified the corresponding location in the field using the GPS device. I then used the data logging capability of the GPS device to record habitat measurements for each point. Finally, I downloaded these data in spreadsheet format to conduct statistical comparisons between the CZ points and the random points.
At each point measured, I recorded information from both the pavements edge and 20 m from the pavement edge, to reflect what an animal would see as it approached the road as well as when it began crossing. At TCP I took measurements from both
62


sides of the road, resulting in a pair of measurements at each point. At VP, I treated the westbound and eastbound alignments separately, locating points to measure independently along each alignment. I recorded data from only the outer edge of the alignment as I deemed it unsafe to cross over to the median-side edge. The variables that I measured and recorded at each point are described in Table 3.3. The line-of-sight (LOSAR, LOSTRa, LOSTRb) and distance-to-cover measurements (D WV, DWV20) were made from the perspective of a medium-sized animal (e.g., coyote, bobcat, mountain lion) as well as from the perspective of a large-sized animal (deer or elk). These two animals eye views of the habitat were simulated by placing a laser rangefinder on a monopod sized to elevate the eyepiece 0.8 or 1.6 m above the ground. Line-of-sight distances and distance-to-cover, as observed through the rangefinder, were then recorded at both monopod heights.
The second type of data I collected in the field was the locations of features along the roadside, including bridges (representing the locations of both underpasses and drainages) and roadside barriers (cliffs, walls, guardrails, Jersey barriers). I used these data to create digital data layers for use with ArcView. I collected these data using the GPS devices setting for recording continuous data along a line. Using a roof-mounted antenna, I drove slowly (20-25 km/h) along each feature of interest and collected positions for the entire length of each feature, then converted the positions into ArcView shapefiles using Trimbles proprietary software package. I used these data layers to measure the distance of CZs to their respective features.
63


Table 3.3 Variables measured in the field at crossing zones (CZs) points and random points for comparison of local-scale habitat charateristics
Variable Measured Abbreviation Notes
At Roadside Line-of-Sight (m) along roadway at 1.6 m, 0.8 m* LOSAR Two LOSAR measurements were taken at each point, one in either direction.
Distance (m) to nearest woody vegetation at 1.6 m, 0.8m* ** DWV Measurements were taken along eight equidistant radii originating from the center of the point being measured, at both heights.
Distance to nearest side road Distance to nearest center of human activity DTSR DTHA Distance to nearest visible side road Distance nearest house, rest area, business, or maintenance depot.
20 m From Roadside Line-of-sight to roadway at 1.6 m and 0.8 m* LOSTRa*** LOSTRb*** View to road from point being measured, categorized as 0 = completely obstructed, 1 = partially obstructed, 2 = unobstructed.
Distance to nearest woody vegetation at 1.6 m, 0.8m* ** DWV20 Eight measurements were taken along eight originating radii emanating from the center of the point, at both heights.
Minimum and maximum slope between roadside and 20 m from road side MINSL, MAXSL Measured in degrees
* These measurements were taken at both 0.8 m above the pavement surface and 1.6 m above the pavement surface
** No values were recorded for radii that intersected rock or dirt. Woody vegetation 70m or farther from the point was all recorded as 70 m.
*** Three observations were taken at each point, one looking straight to the road (LOSTRa), and then two more, looking to the road 45 either side from the line of the first observation. The two side views were averaged (LOSTRb).
In addition to making field measurements of the roadside habitat, I also used aerial photos and existing digital data layers to make some local-scale measurements. I digitized lines representing the forest boundaries from aerial photos, then used that data layer to measure the distance of CZs to the nearest forest edge with ArcView. Additionally, I used the digital vegetation and topographic data layers described in the Landscape-Scale Habitat Measurements section to compare the cover, slope and aspect classes associated with CZs to what was available along the entire roadside.
64


Data Analysis: Identifying Patterns
Descriptive Summary of Track Records
I considered the data in three separate sets, based on study site and snow depth: Trout Creek Pass (TCP), Vail Pass snow-free (VP), and Vail Pass with snow cover (VPS). I calculated the total number of TRs recorded at each site, counted them by activity and species, and created maps depicting the TR locations. I further divided these primary data sets into subsets, as described below, based on the spatial patterns of the TRs within each study site, and landscape characteristics of the study sites. This approach was necessary because visual analysis of the maps indicated that the density of TRs throughout each study site was uneven.
Identifying First Order Patterns
The maps suggested that both first order (large-scale) and second order (small-scale) patterns were present. Because variations in first order patterns can mask or swamp second order patterns, small-scale spatial patterns must be studied over scales at which the first order effects remain homogenous (Baily and Gatrell 1995, p. 91). Therefore, I divided each of the three primary data sets into sub-areas, based on the extent of homogenous first order patterns. I used visual analysis and simple counts of the crossing TRs at each study site to demark their first order patterns. I confirmed the presence of these first order patterns by using SPSS v.10.1 for Windows to perform chi-square tests to determine if the observed distributions between sub-areas of apparent high and low density TRs did indeed differ from an even distribution.
65


Identifying Second Order Patterns (Crossing Zones')
I sought to confirm the presence of second order (small-scale) patterns within the sub-areas by looking for groups of TRs that were closer to one another than expected. For this analysis I used only those TRs that had been classified as crossing in the field so that I could use designated clusters as indicators of CZs. I defined a TR as part of a cluster if the median distance to its n nearest neighbors was less then expected by chance. The median, rather than the mean, distance between a point and its n nearest neighbors was chosen as the metric of comparison for identifying CZs because the distribution of nearest-neighbor distances was skewed to die right. The median is less influenced by outliers, providing a more conservative estimate of the datas central tendency.
Because first order patterns affect the expression of second order patterns, the number of nearest neighbors I considered varied with the total number of TRs in each sub-area. For sub-areas with fewer then 100 TR, I considered the median distances of a point to its three nearest neighbors. For sub-areas with 101 to 199 TR, 1 considered the median distances of a point to its five nearest neighbors, and for sub-areas with 200 to 299 TR, I considered the median distances of a point to its seven nearest neighbors. I defined an n nearest-neighbors distance smaller than three standard deviations from the expected median distance as not occurring by chance, and therefore, an indication of clustering.
To determine the expected nearest neighbors distance for each data set, I used a Monte Carlo (Besag and Diggle, 1977) approach. I distributed points randomly along a line representing the roadway of interest to simulate a possible distribution of crossing TRs along it, and then measured the n-nearest-neighbors distance for each point. I used 1000 simulations to generate the expected distribution of nearest-
66


neighbors distances. A script written for ArcView in the Avenue programming language (Martin, 2001) was used to perform all simulations automatically, based on user-defined input parameters. Using this script, I specified a line, the number of points to use, the number of nearest neighbors to measure, and the number of simulations to run. The script then automatically performed the number of simulations requested, calculated the mean, median, maximum, and minimum n-nearest-neighbors distance of each simulation, and stored these values in a spread sheet. After the simulations were completed, I exported the spreadsheet to Excel and calculated summary statistics of each value.
Measuring Crossing Zones
After identifying the TRs with n closer-than-expected nearest neighbors on a map of all crossing TRs, I buffered them with a radius the length of the expected median w-nearest-neighbors distance, and dissolved the boundaries of all overlapping buffers. I chose this buffer size as it should include all the points that contributed to a given TRs nearest neighbor measurement, on average. I designated all crossing TRs contained within each buffer as part of that CZ (Figure 3.7). I then measured the length of each CZ as the distance between the two outer-most TRs in ArcView. They are accurate to + 10 m. A group of TRs was only considered a CZ if it contained at least as many TRs as were used in that data sets nearest neighbors calculation (e.g., three nearest neighbors required considering four TRs). Finally, I inspected the TRs that comprised a CZ to determine if they represented independent events. TRs of the same species recorded the same day and within 50 m of each other were not considered to be independent.
67


Underpass Use
I calculated the frequency of underpass use in two different ways. First, I compared the number of times at least one set of tracks, of any species, was recorded as passing through to the total number of times the structure was checked for tracks. Although it was common for me to find multiple track sets each time I checked an underpass, I could not determine if one animal passing through an underpass many times or many animals passing through once created the multi-track track sets. Because a group of n animals crossing together may be regarded as one crossing event, rather then n independent events, I grouped these data.
However, because the first approach lumps data, important information about species use is masked. Therefore, I also compared the number of times at least one individual, by species, was recorded as passing through to the number of times the structure was checked for tracks. For example, if I checked an underpass and recorded a bobcat and three rabbits as passing through, I counted the underpass as used once by bobcat and once by rabbit. Using this counting method, it is possible that I could record more through passes then the number of times the underpass was checked. Therefore, I report the frequency of use calculated with this approach as a ratio. Finally, at VPS I also counted the total number of animals passing through the underpasses. Snow is an excellent tracking medium, and few animals are present in the Vail Pass area when the snow is deep, making it possible to accurately count every trail.
68


Data Analysis: Quantifying Relationships
Quantifying the Relationship of First Order Patterns to Landscape Structure
As detailed previously, I split each of the three study sites into two sub-areas, based on the first order patterns of TR density along the roadside. I evaluated the influence of the surrounding landscape on this first order pattern of TR distribution by comparing the landscape structure of the each study sites two sub-areas.
Metrics of landscape structure that I considered included the average size and ESR of patches defined by different classes of cover type, slope, or aspect in each study site sub-area. I used the Arc View extension Patch Analyst to count the total number of patches created by variations in cover, slope, or aspect, as well as to calculate the area of each patch and the length of its perimeter. I defined a patch as a contiguous area comprised of a consistent class. Although some of the samples I compared were neither normal nor homogeneous in their distributions, I chose a two-sample t-test (SPSS 10.1) to compare the mean values of each variable. SPSS offers a correction for heterogeneous distributions, and my samples size were very large, relaxing the need for normal distributions.
Quantifying the Relationship of Second Order Patterns to Local-Scale Features
The approach I used to quantify the local-scale habitat characteristics associated with crossing zones varied according to type of feature under consideration and the datas source. For variables measured in the field (listed in Table 3.3), I calculated the average value of each variable measured within CZs and the average value measured at random points. I then compared these two values to determine if there was a difference between CZs and random locations. Because these data were non-

69


normal and often had significantly different variances, I used the non-parametric Mann-Whitney U test (SPSS 10.1.0 for Windows) to determine if the means differed from one another.
For the cover, slope, and aspect variables measured on GIS data layers, I used the ArcView extension Patch Analyst to calculate the total area of each cover, slope, and aspect class within 100 m of the CZs and within 100 m of the entire roadway. I used Excel spreadsheet functions to run a chi-square test to compare the proportion of cover, slope, and aspect classes contained in each of the paired data sets.
I also used a GIS-based approach to evaluate if the distribution of CZs in relationship to drainages, roadside barriers, and the forest boundary deviated from what would be expected. I used scripts written for ArcView (Martin 2001) in the Avenue programming language to measure the actual distance of each TR to the feature of interest, and to implement a Monte Carlo simulation (n = 1000) to generate an expected distance for comparison. A single simulation consisted of distributing points randomly along a line representing the road, measuring the distance of all points to the nearest feature of interest, then calculating the average point-to-feature distance. The number of points used corresponded to the number of TRs that made up the CZs in the study site sub-area being evaluated.
The script carried out the Monte Carlo simulations automatically, based on user-defined input parameters, including ArcView shapefiles representing the feature of interest and the roadway, the number of points to use, and the number of simulations to run. The script then executed the number of simulations requested, calculated the mean, median, maximum, and minimum distance, and stored these values in a spreadsheet. After the simulations were complete I exported the spreadsheet to Excel and calculated the summary statistics of each value. I used a
70


two-sample t-test, implemented with Excel spreadsheet functions, to compare the actual mean TR-to-feature distance to the expected mean. If the two means were not significantly different at a = 0.05,1 concluded that the TRs were randomly distributed in relationship to the features of interest. Additionally, the script counted the number of points in each simulation that were placed adjacent to a barrier so I was able to calculate an expected value. I compared the proportion of TRs that were expected to be adjacent to a barrier to the actual proportion of TRs located adjacent to a barrier with a chi-square test.
To examine if the presence of underpasses had any effect on the locations where animals crossed at-grade, I graphed the relationship between meters of underpass along a kilometer of highway against the number of crossing TRs along that kilometer of highway. I estimated both values using a moving windows analysis. Each window along the roadway was one km in length and was shifted in 100 m increments.
Quantifying Animal Abundance
The small data sets from the off-road transect collected at both TCP and VPS do not provide a statistically valid sample to evaluate and compare animal densities. Likewise, the gross scale of the CDOW data does not provide an accurate estimation of the animal populations at the roadside specifically. Therefore, I summarized these data by simply counting the number of tracks observed and calculating the average number of tracks/transect at each study site, and reporting the CDOW deer and elk population data.
71


Summary
To determine if the locations where animals cross the highway are different from random locations, I chose variables for comparison by reviewing the current literature and considering which roadway and roadside features highway designers control. To collect data, I chose two study areas in the Southern Rocky Mountains. First, I recorded where wild animals crossed the road. Next, I measured characteristics of the habitat both directly in the field and from digital data layers that were created from field measurements or remote photography. Depending on the type of data collected, I analyzed the data by comparing average values, using a Monte Carlo approach to generate an expected distribution to compare to actual distributions, or by comparing used habitat to available habitat. I also collected data about underpass use and summarized it with simple counts.
72


CHAPTER 4
WILDLIFE HIGHWAY-CROSSING ZONE INVESTIGATION: FINDINGS
Introduction
In the following chapter I report the results of my data collection and analysis efforts. The discussion is organized according to the order followed in Chapter 3. Simple summaries of the track record (TR) data are reported first. Results of the analyses that correlated first order TR patterns to characteristics of the surrounding landscape are followed by results of the analyses that correlated second order TR patterns to characteristics of the roadside. Finally, I summarize the underpass use and the animal abundance trends at all three study areas.
Descriptive Summary of Tracks
I collected track data 130 times at Trout Creek Pass (TCP) between 28 January 2000 and 4 July 2001, recording a total of 535 TRs, representing 832 individual animals. I conducted a total of 91 tracking sessions when there was no snow cover at Vail Pass (VP), comprised of 40 sessions during 2000 and 51 sessions during 2001.1 recorded a total of 778 TRs, representing 1155 individuals, at VP. When there was snow on the ground at Vail Pass (VPS) I collected track data on 18 occasions during 2000/01 and on 30 occasions during 2001/02, for a total of 48 snow tracking sessions, and I recorded a total of 771 TRs, representing 978 individuals. The TRs are summarized by species and activity for each study site in tables 4.1,4.2, and 4.3.
73


Table 4.1 Summary of TRs by species and travel at TCP
Species Crossing (% crossing, by species) Not Crossing (% not crossing, by species) Total (% of all TRs, by species)
Mule Deer 219(53.0%) 194(47.0%) 413(77.2%)
Elk 40 (71.4 %) 16 (29.6 %) 56 (10.5 %)
Coyote 10 (27.8 %) 26(71.2%) 36 (6.7 %)
Rabbits/Hares 7(70.0%) 3(30.0%) 10(1.9%)
Fox 1 (25.0 %) 3 (75.0 %) 4 (0.7 %)
Mountain Lion 1(100%) - 1 (0.2 %)
Bobcat - 1 (100.0 %) 1(0.2 %)
Other 2(40.0%) 3(60.0%) 5 (0.9 %)
Unknown - 7 (100.0 %) 7 (1.3 %)
Total 278 257 535
Table 4.2 Summary of TRs by species and travel at VP
Species Crossing (% crossing, by species) Not Crossing (% not crossing, by species) Total (% of all TRs, by species)
Mule Deer 191 (34.9 %) 284 (65.1 %) 475 (61.0 %)
Elk 117(43.2%) 154 (56.8 %) 271 (34.8%)
Coyote 8 (28.6 %) 20(71.4%) 28 (3.6 %)
Mountain Lion 1 (100.0 %) - 1 (> 0.1 %)
Moose 1 (100.0 %) - 1 (> 0.1 %)
Other - 1 (100.0%) 1(>0.1%)
Unknown 1 (100.0 %) - 1 (>0.1 %)
Total 319 459 778
Table 4.3 Summary of TRs by species and travel at VPS
Species Crossing (% crossing, by species) Not Crossing (% not crossing, by species) Total (% of all TRs, by specie
Coyote 433 (74.5 %) 148 (25.5 %) 581 (75.3 %)
American marten 12 (35.3 %) 22 (64.7 %) 34 (4.4 %)
Weasel species 23(45.1 %) 28 (54.9 %) 51 (6.6%)
Snowshoe hare 39 (76.5 %) 51 (23.5 %) 90(11.7%)
Red fox - 2 (100.0 %) 2 (0.3 %)
Elk - 8 (100.0 %) 8(1.0%)
Total 507 264 771
The species recorded most often at TCP and VP was mule deer (77.2 and 61.0 %, respectively), and I also recorded a substantial number of elk at VP (34.8 %). Coyotes were the species recorded most commonly at VPS (75.3 %). The proportion of tracks indicating crossing varied among the three study sites, as did the animal species that crossed most frequently. Animals were most likely to cross
74


at VPS (65.7 % of all tracks) and least likely to cross at VP (41.0 %). The crossing rate at TCP was intermediate (52.0 %). The crossing rate at TCP was not significantly different from either VP (x2 = 2.289, p > 0.10) or VPS(^ = 3.669, p > 0.05), but animals were significantly more likely to cross the highway at VPS, as compared with VP (%2 = 9.286, p < 0.05).
Distribution of Tracks
Animals are more likely to cross highways at certain locations at both the landscape and the local scale. Quantitative as well as visual analyses of the patterns created by the distribution of track records (TRs) along the roadside indicated that both first order and second order clustering existed at all three study sites. I interpreted the second order TR clusters as indicators of locations where animals preferred to cross the road, and designated such locations as crossing zones (CZs). The results of my spatial pattern analysis of the TRs are described in detail below.
First Order Patterns
Mapping the locations of crossing TRs within their respective study sites revealed that crossing TRs were not evenly distributed throughout any of the study sites. At each of the three sites, I observed more crossing TRs located in a definable sub-area of the site. At TCP I recorded far fewer TRs along the portion of US 24 located north of where Trout Creek intersects Hie highway. The resulting low- and high-density TR areas corresponded with differences in both topography and land cover north and south of Trout Creek, and I divided the data into two sub-areas, north (MP 221.5-226.0) and south (MP 216.0-221.5), accordingly (Figure 4.1).
At VP, I recorded less than half as many TRs on the west side of the Pass than the east side. As discussed in Chapter 3, the design of the highway differs substantially between the two sides of the Pass. Therefore, the Pass was a reasonable dividing
75


line between the differing first order patterns, and I subdivided the data into two sub-areas, east (MP 190.0-195.0) and west (MP 183.0-190.0; Figure 4.2).
When snow is present in the Vail Pass area, large, steep snowdrifts along the roadside are created throughout the study site by snowplows. These drifts masked some of the structural differences of the roadway between the east and west sides of the Pass, and it became less of a natural dividing line. Additionally, the distribution of TRs when snow was on the ground was consistent throughout the VP study area, except for the 2.5 miles of roadway closest to Copper Mountain Resort ski area (CMR). During the months with snow cover I recorded a clearly disproportionate number (75.0 percent of total) of TRs in the portion of the study area closest to CMR. Therefore, I divided VPS into two sub-areas, CMR (MP 192.5-195.0) and Not CMR (MP 183.0-192.5), based on the location of the resort (Figure 4.3).
At all three sites the first order patterns, which were apparent when I mapped the locations of all TRs, were mimicked when I mapped crossing TRs only. Additional quantitative analysis supported my initial impressions. A simple count revealed that at TCP nearly 80 % of crossing TRs were located in the southern half of the study area, at VP 68.5 % of crossings TRs were located on the east side of Vail Pass, and at VPS 68.4 % of crossing TRs were adjacent to CMR. Chi-square tests indicated that the proportion of crossing TRs located in the sub-areas departed significantly from either an even or a random distribution, based on the linear distance of the roadway at each study site sub-area (Table 4.4).
76


Figure 4.1 Locations of all crossing TRs recorded at TCP. Note the higher density
of TRs in the south end of the study area.
77


Figure 4.2 Locations of all crossing TRs recorded VP. Note the higher densities of
tracks on the east side of the pass.
78


Eagle County
N
A
Figure 4.3 The locations of all crossing TRs recorded VPS. Note the higher density
of tracks near CMR.
79


Table 4.4 Distribution of crossing TRs within the sub-areas of each study site
Study Site Sub-Area Number of Crossing TRs Linear Length of Sub-Area TRs/km Chi-Square Comparison
TCP South 219 8.9 km 24.6 X2= 19.13, p< 0.00
North 59 7.3 km 8.1
VP East 227 8.8 km 25.8 X2= 80.20, p< 0.00
West 92 11.3 km 8.1
VPS CMR 401 4.0 km 100.2 X2= 214.06, p< 0.00
Not CMR 106 15.4 km 6.7
Second Order Patterns
Visual analysis of the mapped crossing TRs in the sub-areas of each study site suggested that additional small-scale, or second order, clustering of TRs was present within the first order clustering discussed above (Figures 4.1,4.2, and 4.3). Nearest neighbors analyses confirmed my impression of second order patterns at all three study sites. At TCP south the nearest neighbors analysis indicated that 60.4 % of TRs were more clustered than would be expected by chance. At TCP north 58.6 % of TRs were more clustered than expected by chance. At VP and VPS, I analyzed clustering separately in the west- and eastbound lanes. At VP, TRs recorded along the westbound lanes were more likely to be clustered together than those recorded along the eastbound lanes. This was true at both VP east (71.8 versus 47.8 %) and VP west (66.7 versus 34.4 %) of the Pass. At VPS, 42.9 of the TRs in the westbound lanes and 57.1 % of TRs in the eastbound lanes were closer to one another than expected adjacent to CMR. In the rest of the study area, 63.9 % of TRs along the westbound lanes and 72.9 % of TR along the eastbound lanes were more clustered than expected by chance.
80


Crossing Zones
I identified CZs based on the second order patterns of crossing TRs. I interpreted groups of crossing TRs that were more clustered than expected by chance as an indication that animals focused crossing activity along that stretch of highway. At TCP north I identified five distinct CZs (Figure 4.4). A sixth CZ was discarded because the TRs that defined it were all the result of crossing by a single herd of elk and could not be considered independent events. The identified CZs ranged from about 80 m to 300 m in length. Distance between them ranged from 240 m to 2630 m. At TCP south, I identified 10 CZs (Figure 4.5). They ranged from about 30 m to 600 m in length, and the distance between them ranged from 200 m to 1120 m.
At VP east, I identified five distinct CZs along the westbound lanes and four along the eastbound lanes (Figure 4.6). These CZs ranged in length from 100 to 760 m and the intervals between them ranged from 140 to 1510 m in length. I dropped three additional CZs along the westbound lanes from consideration because they were defined by four or fewer TRs, and in each case at least three of the TRs were from the same species, recorded during the same data collection session, and could not be considered independent events. CZs along the west- and eastbound lanes were not strongly aligned with one another. At VP west, I identified five distinct CZs along the westbound lanes and three along the eastbound lanes (Figure 4.7). Crossing zones on this side of the pass ranged in length from 50 to 490 m in length and the intervals between them ranged from 650 to 2480 m in length. I dropped one CZ along the westbound lanes because all but one of its identifying TRs were recorded during the same data collection session and were from the same species. All three of the CZs identified along the eastbound lanes are strongly aligned with CZs along the westbound lanes.
81


Chaffee County
US 24
N
A
Figure 4.4 Locations of the crossing TRs that make up the CZs in the north end of TCP.
82


Figure 4.5 The locations of crossing TRs that make up the CZs in the south end of TCP
83


Figure 4.6 Locations of crossing TRs that make up the CZs on the east side of VP.
84


Figure 4.7 Location of crossing TRs that make up CZs on the west side of VP.
85


Full Text

PAGE 1

CHARACTERISTICS OF WILDLIFE HIGHWAY-CROSSING ZONES AND APPLYING THIS INFORMATION TO REDUCE WILDLIFEIHIGHW A Y CONFLICTS by Sarah Barnum B.S., University ofVennont, 1988 M.S., Utah State University, l994 A thesis submitted to the University of Colorado at Denver in partial fulfillment of the requirements for the degree of Doctor of Philosophy Design and Planning 2003

PAGE 2

This dissertation for the Doctor of Philosophy Degree by Sarah A. Barnum has been approved by Fahriye Sancar David Armstrong James Huff

PAGE 3

Barnum, Sarah A. (Ph.D., Design and Planning) Characteristics of Wildlife Highway-Crossing Zones and Applying This Information to Reduce Wildlife/Highway Conflicts Dissertation directed by Professor Willem K. T. Van Vliet ABSTRACT Negative impacts of highways on wildlife include highway-induced habitat loss and alterations, habitat fragmentation, direct mortality, and disturbance. Although the list of documented impacts is substantial, less is known about the mechanics of wildlife/highway interactions. However, a review of literature pertaining directly to where wild animals interact with roads suggested that at grade crossings do not occur at random locations and that variables from both the local and the landscape scale play a role in the location of wildlife/highway interactions. To verify if the locations where animals cross highways are different from random locations, I recorded where wild animals crossed the road and measured characteristics of the surrounding habitat at two locations in the Southern Rocky Mountains. Depending on the type of data collected, I analyzed it by comparing average values, using a Monte Carlo approach to compare expected with actual distributions, or by directly comparing used with available habitat. I also collected data about underpass use and summarized it with simple counts. A qualitative assessment of my results indicated that crossing zones are related to variables from both the landscape and the local scale. Significant variables iii

PAGE 4

included features from both the habitat and roadway. At the large scale, the most important features were cover type composition, slope, and slope complexity of landscape surrounding the highway. At the local scale, the most important features were the location of and distance to roadside barriers, the location of drainages, and the distance from the road to the forest edge. These variables provide an accessible source of information to improve the practice of reducing wildlife/highway conflicts. Conflict reduction should include identifying conflict locations at both a landscape and a local scale, considering conflict locations when choosing the best strategies and locations for mitigation, and integrating this process into highway project planning up front. Additional data about wildlife/highway interactions as well as coordination among highway planners, conservation planners, and entities that affect land use, will improve the practice of reducing wildlife/highway conflict in the future. This abstract accurately represents the contents of the candidate's dissertation. I recommend its publication. Signe Will em K. T. Van Vliet iv

PAGE 5

ACKNOWLEDGEMENTS My thanks to everyone who helped make it possible for me to earn my Doctor of Philosophy. I extend special thanks to: My advisor and all my committee members for their advise and support while I worked on this project. The Colorado Department of Transportation for their generous funding and the freedom to conduct my research as I saw fit. My field assistants, Eli Wostl, Christiana Manville, and Marcus Pacheco, who always worked hard and did a great job.

PAGE 6

CONTENTS Figures ....................................................................................................... xii Tables ....................................................................................................... xiv CHAPTER 1. REDUCING WILDLIFEIHIGHW A Y CONFLICTS ............................. I Introduction ..................................................................................... I An Overview of Wildlife/Highway Conflicts ................................. 2 Habitat Fragmentation ......................................................... 3 Habitat Loss and Alteration ................................................. 4 Direct Mortality ................................................................... 6 Disturbance .......................................................................... 7 A Historical Perspective of Wildlife/Highway Conflicts ................ 8 Traditional Highway Planning and Wildlife ....................... 8 Traditional Conservation Planning and Highways ............ I4 The Current Institutional Setting for Reducing Wildlife/Highway Conflicts ........................................................................................ I8 The Regulatory Nexus to Consider Wildlife in Highway Planning ............................................................................. 20 Highways and the Ecosystem Approach to Conservation Planning ............................................................................. 23 VI

PAGE 7

The Current Practice ofReducing Wildlife/Highway Conflicts ............................................................................ 25 Summary ....................................................................................... 27 2. A REVIEW OF RESEARCH RELATED TO IDENTIFYING WILDLIFE IDGHW A Y-CROSSING ZONES ..................................... 28 Introduction ................................................................................... 28 Roadside Habitat Use .................................................................... 30 Animal/Vehicle Collision Locations Along Highways ................. 31 Use of Highway-Crossing Structures by Wildlife ......................... 33 Locations of At-Grade Highway Crossing by Wildlife ................ 35 Summary ....................................................................................... 38 3. IDENTIFYING WILDLIFE IDGHWA Y CROSSING ZONES IN THE COLORADO ROCKIES: RESEARCH BACKGROUND .......... 39 Introduction ................................................................................... 3 9 Overview ........................................................................... 39 The Research Approach .................................................... 40 Relevance of the Measured Variables ........................................... 41 Landscape Scale Variables ................................................ 42 Local-Scale Variables ....................................................... 44 Study Site Descriptions ................................................................. 45 Trout Creek Pass ............................................................... 46 VII

PAGE 8

Vail Pass ............................................................................ 49 Data Collection Methods: Tracking .............................................. 54 Standard Tracking Methods .............................................. 54 Standard Underpass Monitoring ........................................ 56 Snow Tracking Methods ................................................... 57 Animal Abundance ............................................................ 59 Data Collection Methods: Habitat Measurement .......................... 60 Landscape Scale Habitat Measurements ........................... 60 Local Scale Habitat Measurements ................................... 62 Data Analysis: Identifying Patterns ............................................... 64 Descriptive Summary of Track Records ........................... 65 Identifying First Order Patterns ......................................... 65 Identifying Second Order Patterns (Crossing Zones) ........ 65 Measuring Crossing Zones ................................................ 67 Underpass Use ................................................................... 67 Data Analysis: Quantifying Relationships .................................... 68 Quantifying the Relationship of First Order Patterns to Landscape Structure .......................................................... 69 Quantifying the Relationship of Second Order Patterns to Local Scale Features .......................................................... 69 Quantifying Animal Abundance ....................................... 71 Summary ....................................................................................... 72 viii

PAGE 9

4. WILDLIFE HIGHWAY-CROSSING ZONE INVESTIGATION: FINDINGS ........................................................................................... 73 Introduction ................................................................................... 73 Descriptive Summary of Tracks .................................................... 73 Distribution of Tracks ................................................................... 75 First Order Patterns ........................................................... 75 Second Order Patterns ....................................................... 80 Crossing Zones .................................................................. 81 The Relationship of First Order Patterns to Landscape-Scale Features ......................................................................................... 86 Composition ...................................................................... 89 Complexity .. ; ..................................................................... 90 Other Influences from the Landscape Scale ...................... 93 The Relationship of second Order Patterns to Local Scale Features ... ............. ....................................................... 94 Features Measured Directly in the Field ........................... 94 Features Measured from GIS Data Layers ........................ 96 Underpass Use ............................................................................. 103 Trout Creek Pass ............................................................. 104 Vail Pass .......................................................................... 106 Vail Pass Snow ................................................................ 108 ix

PAGE 10

Animal Abundance ...................................................................... II 0 Trout Creek Pass ............................................................. 110 Vail Pass .......................................................................... Ill Vail Pass Snow ................................................................ Ill 5. VARIABLES TIIAT IDENTIFY WILDLIFE HIGHWAY CROSSING ZONES: DISCUSSION .................................................. 113 Introduction .... ; ............................................................................ 113 Descriptive Summary of Tracks and Their Distribution ............. ll4 First Order Patterns and Landscape-Scale Variables .................. 115 Composition and Complexity .......................................... 115 Other Landscape Scale Influences .................................. 117 The Relationship of Second Order Patterns to Local Scale Features "; ... ............................................................................ 120 Variables Measured Directly in the Field ....................... 120 Features Measured from GIS Data Layers ...................... l22 Underpass Use ............................................................................. 126 Trout Creek Pass ............................................................. 126 Vail Pass .......................................................................... 127 Vail Pass Snow ................................................................ l27 Summary ..................................................................................... 128 X

PAGE 11

6. APPLYING THE RESEARCH RESULTS TO REDUCE WILDLIFEIIDGHW A Y CONFLICTS ............................................... 129 Introduction ................................................................................. 129 Are the Results ofthe Study Useful? .............................. 130 Strategies for Identifying Wildlife/Highway Conflict Locations ......................................................................... 131 A Strategy for Identifying Conflict Zones ...................... 131 A Strategy for Identifying Crossing Zones ..................... 135 Approaches for Reducing Wildlife/Highway Conflicts .............. 137 Identify and Avoid High Conflict Locations ................... 138 Managing the Surrounding Landscape to Reduce Conflicts .......................................................................... 139 Design Based Approaches for Reducing Conflicts ......... 140 Integrating Mitigation Planning into Highway Planning ............ 143 Improving the Practice of Reducing Wildlife/Highway Conflicts ................................................................. : ..................... 144 Summary ..................................................................................... 147 APPENDIX A. ABBREVIATIONS USED IN THE TEXT ....................................... 148 BIBLOGRAPHY ................................................................................................ 149 XI

PAGE 12

FIGURES Figure 3.1 The location of the Trout Creek Pass study site (TCP) in Chaffee County, Co lorado ........................................................................................... 4 7 3.2 Detail of the Trout Creek Pass area, location of TCP .................................... 48 3.3 Location of the Vail Pass (VP) and Vail Pass Snow (VPS) study sites straddling Eagle and Summit Counties in Colorado ...................................... 50 3.4 Detail of the Vail Pass area, location of both VP and VPS, including the location of the Copper Mountain Resort (CMR) ........................................... 51 3.5 The locations of all barriers and underpasses on 1-70 at Vail Pass ................ 53 4.1 Locations of all crossing TRs recorded at TCP .............................................. 77 4.2 Locations of all crossing TRs recorded at VP ................................................ 78 4.3 Locations of all crossing TRs recorded at VPS .............................................. 79 4.4 Locations of the crossing TRs that make up the CZs in the north end of TCP ............................................................................................................ 82 4.5 Locations of the crossing TRs that make up the CZs in the south end of TCP ............................................................................................................ 83 4.6 Locations of the crossing TRs that make up the CZs on the east side of VP ............................................................................................................. 84 4.7 Locations of the crossing TRs that make up the CZs on the west side of VP ............................................................................................................. 85 4.8 Locations of the crossing TRs that make up the CZs in the sub-area of VPS adjacent to CMR ................................................................................... 87 vi

PAGE 13

4.9 Locations of the crossing TRs that make up the CZs in the sub-area of VPS not adjacent to CMR ............................................................................. : 88 4.10 Detail of the summer crossing TRs that made up CZs, in relationship to barriers and underpasses, on the east side of VP .......................................... 98 4.11 Detail of the winter crossing TRs that made up CZs, in relationship to barriers and underpasses, on the east side ofVP .......................................... 99 4.12 Detail of the summer crossing TRs that made up CZs, in relationship to barriers and underpasses, on the west side ofVP ....................... .............. 100 4.13 Detail of the winter crossing TRs that made up CZs, in relationship to barriers and underpasses, on the west side of VP ...................................... 101 6.1 A framework for mitigating wildlife/highway conflicts along existing highways : ............ ; ..... ; .......... .................. ..................................................... 133 Vll

PAGE 14

TABLES Table 1.1 Environmental statutes with potentially extensive effects on the placement and design ofhighways in the u.s ................................................................. 22 2.1 A summary .of variables demonstrated to be important in determining the locations where animals interact with roads and highways ........................... 29 3.1 Habitat and roadway variables measured to compare crossing zone (CZ) locations to random locations ..................................................................... 42 3.2 Definitions used to divide cover, slope, and aspect classses on digital maps of the three study sites .......... : .................... : ..................... ; ................... 61 3.3 Variables measured in the field at crossing zone (CZ) points and random points for comparison of local scale habitat characterisitcs .......................... 63 4.1 Summary ofTRs by species and travel at TCP .............................................. 74 4.2 Summary ofTRs by species and travel at VP ................................................ 74 4.3 Summary ofTRs by species and travel atVPS .............................................. 74 4.4 Distribution of crossing TRs within the sub-areas of each study site ............ 80 4.5 Comparisons of sub-area pairs by composition ............................................. 90 4.6 Comparison of landscape metrics associated with first order patterns .......... 91 4.7 Average values of measurements taken at CZ and random point locations ... 95 4.8 Results of chi-square tests comparing the distribution of cover, slope, and aspect classes within 100m of the CZs to what is available throughout the highway corridor within 100 m of the roadside ............................................. 97 viii

PAGE 15

4.9 Actual distances compared with the expected distances ofCZs to barrier ends and the results of the chi-square tests comparing the actual with the expected number of TRs located mid-barrier ............................................... 102 4.10 Actual distances compared with expected distances ofCZs to the nearest drainage that intersects with the road ............................................. 103 4.11 Relationship of CZs to the forest edge ....................................................... 103 4.12 Characteristics of monitored underpasses at TCP and the number of times at least one through-pass by at least one medium-or large-sized mammal was recorded ................................................................................. 105 4.13 Number of times at least on individual of a species used each underpass at TCP ........................................................................................ 105 4.14 Characteristicsofmonitored underpasses at VP and the number of times at least one.through-pass by at least one medium-or large-sized maJDDlal was recorded ........ ....................................................................... 107 4.15 Number of times at least on individual of a species used each underpass at VP ......................................................................................... 107 4.16. Characteristics underpasses at VPS and the number of times at least one through-paSs by at least one medium-or large-sized m3JlliD.al was recorded ................................................................................ 109 4.17 Number of times at least on individual of a species used each underpass at VPS .......................................... ............................................. 109 4.18 The total number of animals that used the underpasses at VPS ................. 109 4.19 The number of animal trails/transect recorded in the snow along offroad transects at TCP and VPS, by species ................................................ 111 5.1 Results ofthe Mann-Whitney test comparing means of local-scale habitat measUrements taken at point locations within CZs and random point locations ....................................................... ...................................... 120 IX

PAGE 16

CHAPTER 1 REDUCING WILDLIFE/IllGHWA Y CONFLICTS Introduction Roads arguably create human's in.ost pervasive physical impact on landscapes throughout the world (Forman, 1998). Yet, despite their extensive character and wide array of negative effects on nature, biodiversity/road conflicts have only recently come under scrutiny (Hordequin, 2000). As the list of documented road based impacts to wildlife species and their habitat expands, it is becoming apparent that highway and wildlife conservation planners must cooperate in order to reduce these impacts. However, it is equally apparent that additional information about highway/wildlife interactions is required to create a successful collaboration (Forman, 1998). For example, highway a.I)d conservation planners who wish to reduce roadkill and/or road-caused habitat fragmentation need baseline information about how wild animals interact with roadway. Both these conflicts are linked to the ability of animals to cross highways. Therefore, information about preferred crossing locations would be immensely useful for designing highways that avoid or mitigate high conflict areas, and allow animals to move safely and freely across the roadway. With the need to generate and apply this type of information in mind, the goal of my research wastwo-fold. First, I wanted to determine if the locations where wild animals cross highways are correlated to defmable characteristics of the 1

PAGE 17

surrounding environment and the roadway itself. Then, I was interested in how this type of information could inform the highway design process to wildlife/highway conflicts. I hypothesized that: Wild animals do not cross highways at random and the locations that they use can be correlated to features from the roadside and the surrounding habitat. The identification of likely crossing locations is an accessible process that can be incorporated efficiently into highway design to reduce wildlife/highway conflicts. To place the need to generate and apply this type of information in context, I begin this Chapter by reviewing what is known about of wildlife/highway conflicts. Next I give a broad overview of the development ofboth highway and planning and the historical conditions that prevented these two sets of specialists from generating and using knowledge about wildlife/highway conflicts to reduce impacts. I conclude by examining the existing institutional settings that form a basis for highway and conservation professionals to join forces and create highways designed with wildlife in mind. I then briefly examine the current practice of mitigating highway impacts on wildlife. An Overview of Wildlife/Highway Conflicts Impacts of modem highways on the natural environment in general and on wildlife populations in particular are extensive and relatively well documented (Bennet, 1991; Evink et al., 1996, 1998; Spellerburg, 1998; Trombulak and Frissell, 2000). There are four broad categories of impacts that may occur to wildlife as a result of highway projects including 1) habitat fragmentation 2) habitat loss and alteration, 2

PAGE 18

3) direct mortality, and 4) disturbance. These four impact types can take different forms and may be direct or indirect. Habitat Fragmentation Highwaysfragment habitats when they create a physical or behavioral barrier to animal movement. This barrier effect occurs when animals avoid habitat near roads, are physically unable to cross a road, or are killed while attempting to cross. The-consequences ofthese effects are restrictions on daily movements between resource areas as well as on long distance dispersal and seasonal migration. These types of impacts may have significant effects at the population -level. For example, Sweanor et al., (1999) studied metapopulation dynamics of cougars (Felis concolor) in New Mexico. They documented a loss of dispersal capability by cougars across an expanded highway. Because one side of the highway acts as a source habitat for a smaller block of habitat on the other side, they concluded that the cougar population in the smaller block of habitat is unlikely to persist over time. The degree to which a barrier contributes to habitat fragmentation is scale dependent. For example, at a local scale, if a road passes through habitat that is not preferred by a species, there would appear to be no barrier effect. However, if that area lies between two areas of preferred habitat, the road may act as a barrier at the landscape scale. An animal's perspective also dictates the magnitude ofbarrier effects. While most deer most probably easily cross a roadway 20 m in width, it may be an insurmountable barrier for small mammals or species that are behaviorally averse to entering open areas. No quantitative data exist regarding how a road's design regulates its barrier effect. However, it is logical to assume that barrier effects increase for all species with increased width and the addition of 3

PAGE 19

retaining walls, fences, raised medians, guard rails, and significant increases in volume and/or speed of traffic. Habitat Loss and Alteration Direct habitat loss occurs when a highway project's footprint removes natural cover. In addition to effects that result from a simple loss of cover, the impact of habitat loss depends on the overall availability of a habitat type in the project area and its role in an animal's life cycle. Loss of abundant habitat types may be relatively inconsequential; loss of a rare habitat type will have a greater impact, especially if it provides an essential resource for a species (e.g., nestingldenning sites, important food sources). Other impacts may cause loss ofhabitat indirectly by restricting access. In addition to direct habitat loss, highways also alter the environment in other ways that can be detrimental to wildlife. Replacement of native cover types with non native roadside plantings alters habitat, and cut and fill construction techniques obliterate natural landforms. The physical presence of highways, as well as the cuts and fills, can change both surface and ground water flows. Changes to water flows often cause secondary environmental alterations which in turn can cause major alterations to habitat. Depending on the location of a roadway in the landscape, it may cut some areas off from normal water flows while consequently directing higher peak flows and more intense floods to others. Increased water flows and the accompanying debris and sediments can cause physical alteration of terrestrial and aquatic environments at both the local and landscape scale (Jones et al., 2000). Highways are also widely documented to have significant impacts on water quality due to inputs of heavy metals, salts, and nutrients associated with the roadway 4

PAGE 20

{Trombulak and Frissell. 2000). Changes in hydrological regime and water quality can cause changes in plant dispersal and survival, leading to plant community shifts over time. For example, Findlay and Bourdages (2000) found significant correlations between lower plant diversity and higher densities of roads within Canadian wetlands. Because these types of disruptions affect basic ecological processes, highway impacts may extend far beyond the roadside, affecting an entire ecosystem's function (Forman, 2000). Another impact of highways that can cause profound alterations to natural communities is the role they play in aiding the spread of non-native species and expanding the range of native species into previously unoccupied habitats. Both plant and animal species take advantage of the new pathways created by roads, and these invasive species.often have deleterious effects on resident species that are not adapted to competing with them. Plants species may be spread intentionally as roadside plantings, or unintentional by seedscontainedin mulchesand erosion control coverings. Additionally, the disturbed soils and sunny, roadsides of newly constructed road projects create ideal growing conditions for most pioneer species, many of which are adapted to disperse widely via wind, water, or animal carriers. Animals may follow along roads during dispersal because the artificially smoothed roadside offers easy travel, or they may be lured along roadsides by new food resources created by the roadside plant community. This resource includes the plants themselves, or increased populations of small mammals and birds responding to the new cover type, which in turn attracts predators. Additionally, human users of the road and adjacent areas often leave behind trash that can also be an attractive food source for some animal species. 5

PAGE 21

Direct Mortality Although no data exist quantifying the extent of construction-related wildlife mortality (Trombulak and Frissell, 2000), it is logical to assume that small, local populations of sessile or slow-moving organisms may suffer significant negative consequences from construction. Data documenting mortality due to daily operations of highways, i.e., collisions with vehicles, do exist and cover a wide range species, including mammals, birds, amphibians, and invertebrates (Trombulak: and Frissell, 2000). Roadkill is known to be a limiting factor for some populations of endangered species, including the American crocodile ( Crocodilia americanus; Kushlan, .1988), and the Florida panther (Foster and Humphrey, 1995). Roadkill is also known to have devastating local effects on.small populations. For example, Jones (2000) reported that a road upgrade in Cradle Mountain-Lake St Claire National Park, Tasmania, promoted higher vehicle speeds, resulting in increased rates for eastern quolls (Dasyurus viverrinus) and Tasmanian devils (Sarcophilus laniarius), both ofwhich were subsequently extirpated from the park as a result of this increased mortality. However, although an estimated one million vertebrates are killed daily on America's roads (Defenders, 2002), accurate, current data about the mortality rate and population-level impact on common species are largely unavailable. For example, an estimated minimum one-half to three quarter million deer are annually killed nationwide (Romin and Bissonette,l996a; Hubbard et al., 2000), yet there have been few efforts to document the actual number or rate of deer mortality as a result of collisions with vehicles (Romin and Bissonette, 1996a). 6

PAGE 22

Disturbance Disturbance results from many sources, including construction, day-to-day road operations, and increased human access to the area as a result of road improvements. Possible impacts of disturbance inclu.de direct mortality, temporary avoidance of an area, and permanent abandonment of the surrounding habitat. All these impacts may potentially interrupt activities (e.g., feeding, breeding, travel) essential to survival at both an individual and a species level. Additionally, disturbance may contribute to both habitat loss and fragmentation. The impacts of construction-related disturbance are a function of the species' susceptibility to disturbance, duration of the disturbance, area affected, type of disturbance (e.g., heavy equipment noise versus blasting noise), season, and time of day. Disturbances that last a long time, are loud, unpredictable, arid/or affect large areas will have the greatest impact.' Day-to-day road operations have been shown to cause permanent disturbance effects. Many species are mown to avoid areas of disturbance, thereby reducing or eliminating the habitat value of these areas. Types of disturbance from highway operations inClude noise, visual stimuli, human activity, and pollution. Research (van der Zande et al., 1980; Reijnen et al., 1995; Reijnen et al., 1996) indicates that breeding bird densities are reduced near roads, with the effect being greater for heav}r traffic and reaching farther in open habitats (up to 2000 meters) as compared\vith forested habitats (up to 1500 meters). In Colorado, both mule deer (Odocoileus hemionus) and elk (Cervus elaphus) were shown to avoid areas within 200 meters of a road, with this effect appearing stronger in shrub cover types, as compared with forested habitats (Rost and Baily 1979). Studies also indicate that a variety of carnivores, including grizzly bears ( Ursus horribilis; McLellan and Shackleton, 1988), wolves (Canis lupus; Thiel, 1985; Mech et al., 1988) ail.d bobcats (Lynx rufus; Lovalio and Anderson, 1996) 7

PAGE 23

avoid habitats adjacent to roads. Permanent disturbance effects will contribute to both habitat loss and fragmentation. Road projects also often provide increased human access to previously unused areas. Increased human activity can severely reduce or eliminate the habitat value of an area for many species by eliciting an avoidance response. Shy species, such as most felids, are particularly susceptible to this impact. Increased human presence will also contribute to habitat loss and fragmentation. A Historical Perspective of Wildlife/Highway Conflicts The histories of highway planning and wildlife conservation planning in America share a similar time line. Both disciplines began informally upon the settlement of North America by Europeans, became more formalized in the early 1900's in response to increasing scientific knowledge and public pressures, and then to evolve under these two influences from the 1950's onwards. Both disciplines also have largely ignored one another until very recently. Although highways can have profound effects on wildlife, as discussed above, most planned highways have been designed with almost no consideration for the environment, while conservation theorists and planners made few, if any attempts to comprehensively address these impacts, offer solutions, or otherwise engage highway professionals. I discuss the history of these two disciplines in the U.S. below. Traditional Highway Planning and Wildlife From European settlement until the early 1900's, American highways were not planned, per se. The natural environment, in the form of the landscape and its hydrological and geologi'?al processes, and the needs oflocal.human populations 8

PAGE 24

dictated the location and form of highways in America. These highways usually did not connect point A to point B by the most direct route. Instead, they followed the line of least resistance through the intervening landscape (Lane, 1950). Early highway "construction" consisted simply of clearing vegetation, and perhaps limited grading, filling of wet areas, and constructing bridges across small streams (Patton, 1986). As a result, the environmental impacts of these highways, including impacts to wildlife, were low. These highways did not constitute -major barriers to animals. Habitats were not fragmented except at the most local scales. Landforms were not altered to accommodate the road. Sensitive habitats that were difficult to traverse, such as wetlands and high elevation zones, were avoided. Low traffic speeds meant that roadkill was only an issue for the smallest, slowest moving animals. Beginning in about 1890, the Good Roads Movement blossomed in America. Bicycling clubs, which wanted improved surfaces away from crowded local roads for their new form of recreation, were the main proponents for improved roads initially (Patton, 1986; Kaszynski; 2000), but about this time the automobile appeared on the scene as well. By 1902 over 50 auto clubs had formed nationwide and joined the call for Good Roads (Kaszj.rnski, 2000). In response to these public demands, improved highways were constructed using newly codified engineering principles including surveying, siting, and surfacing techniques, as well as application of geometric design to the shape of the roadbed (USDOT, 1976). The application of these principles can also be considered America's first widespread use of formal highway planning, and it did not include any considerations to reduce impacts to the natural environment. In the early 1900's it was not yet apparent to anyone that roads could cause negative impacts to natural systems. The formal study of ecology was still in its infancy (Kingsland, 1991) and understanding of ecological function was consequently almost non-existent at this time. Roads had 9

PAGE 25

previously caused only the most minimal impacts, and the natural environment was not yet widely viewed as a resource in need of protection. Initially, the form and location of these planned highways did not differ much from traditional rights-of-way, and the alignment and grade remained conditioned on horse drawn traffic (Hewes, 1950). By 1925, however, automobile ownership was common, and highway engineers realized that the consequent increasing average traffic speed called for additional innovations in highway design. Speed-friendly designs required moderate grades and gentle curves qualities that often did not occur along the path of least resistance which most traditional rights-of-way followed through the landscapes. The solution to this problem was to reshape the landscape to fit the needs of the highway. By 1940, standard highway designs reflected an increasing tendency among highway engineers to "break away from conditions previously accepted as determining factors in highway location" (Hewes 1950, p. 327). As the form of the highway departed from the form ofthe landscape, and traffic speeds increased, roadkill became a noticeable phenomenon, as evidenced by three early journal articles. "The Toll of the Automobile" (Stoner, 1925), "Birds and Motor Cars" (Cottam, 1931), and "The Automobile as a Destroyer of Life" (Davis, 1934) are probably the first published accounts of environmental impacts associated with highways. However, these authors focused on the impacts of using the highway, and probably had a very limited audience. Widespread appreciation of environmental impacts caused by highway itself remained elusive at this point, with one exception. This was the visual and aesthetic experience of driving on the highway. 10

PAGE 26

The push to create an efficient American highway system during the early 1900's had focused limited funds on pavement rather than the roadside (Neale, 1950). However, as traffic volumes increased along urban highways, a clutter of billboard advertisements and businesses sprang up along them, creating both a safety hazard and a visual blight. The public, accustomed to "natural" country roads, also found new, unlandscaped rural highways ugly and maintenance issues associated with these untreated roadsides were a significant problem. The bare, erosion-prone cuts and fills which many road projects created made road washouts commonplace (Robinson, 1971 ). Partly in response to these concerns the American Association of State Highway Officials (AASHO) and the Highway Research Board (HRB) joined forces in 1930 to form a committee to study all phases of roadside design, construction, and maintenance. When the Joint Committee disbanded in 1940, AASHO's Committee on Roadside Development took up its work. It focused its attention on the "basic principles of landscape design and practice which should be incorporated in postwar plans for all highway construction." Its fmdings were summarized in a 1943 report that advocated construction of"the complete highway" that would be designed to maximize "the four basic qualities of utility, safety, beauty, and economy" (Neale, 1950, p. 321). The report was, however, merely a recommendation and was never adopted as official policy by either AASHO or the Bureau of Public Roads. Highway project planners were free to interpret or ignore the four basic qualities as they saw fit. Nevertheless, the quality of"beauty" in effect became the yardstick whereby environmental impacts were acknowledged and measured both by highway engineers and highway users. For this assessment, only the environment directly adjacent to the roadside was of interest. During the 1940's, the science of ecology 11

PAGE 27

was still in its infancy and the concept of ecosystems and their interconnections were not yet widely disseminated (Kingsland, 1991 ). Very few people understood that a narrow, linear structure like a highway could have effects on natural systems that extended far beyond the roadside. It was assumed that all the impacts of the highway could simply be constructed away with artful grading and naturalistic plantings that blended with the surrounding landscape. This very limited interpretation of the impacts of highways on natural systems was inadvertently reinforced by the 1956 Highway Act (and all subsequent Highway Acts until the 1990s) that funded construction of the interstate highway system. The Act specified that planners and designers had to follow AASHO's geometric design standards in order to receive funding for their project. These standards had been adopted as part of the Act to ensure highway consistency nationwide (USDOT, 1976). This single design mandate effectively narrowed the focus of highway designers. AASHO's design standards were based on the proposed design speed and projected traffic volume only. No other considerations were mandated, although they were also not forbidden. lnpractice however, the AASHO standards were interpreted to mean that only design speed and traffic volume could be considered, and state highway began planning and designing to those criteria, to the exclusion of all other considerations (Myerson, 2000). Roadside beautification projects continued .to be. incorporated in many projects, but any concern about impacts .beyond the immediate roadside was effectively short circuited by these narrow standards. The "pavement only" emphasis created by the Highway Act funding was further compounded by the elitist attitudes of highway experts and the great pressure they faced from the public and private sectors to get the interstate built. The Good Roads era of the early 1900's had seen the creation of the American highway professional. 12

PAGE 28

Congress intentionally appointed an engineer to lead the first federal road agency, in part because road building was becoming increasingly technical, and in part because it wanted an aura of apolitical expertise associated with federally funded highway projects (Seely, 1987). Thus, from the beginning, highway professionals had both a selfand a public-image of impartial experts applying a complex science that only they understood. By the 1950's, this attitude was deeply entrenched. The engineers and bureaucrats who controlled the highway building process did not consider contributions from non-engineers valid (Patton 1986), and they actively disregarded input from the public and specialists from other disciplines in regards to highway projects. Additionally, by the time construction on the interstate system was initiated in 1956, the existing highway system; constructed mainly before World War II, was hopelessly overburdened. Faced with an overwhelming clamor for better roads, from both the public and commercial interests, the main priority of the state agency highway professionals charged with building the interstate was to construct efficient roadways that maximized user benefits, as quickly as possible (Kaszynski, 2000). The push for efficiency, defmed as more cars, faster and safer, inspired direct alignments from point A to point B. In order to achieve this goal, curves of the landscape were cut, wetlands were filled, streams channalized, and remote mountain passes were paved. Cuts and fills, combined with wider roadways, unnatural surfaces, and higher traffic speeds and volumes created substantial barriers to wildlife movements, and the silt-laden run-off from roadsides had severe impacts to aquatic systems (Mowbray, 1969). However, the impact of all these habitat alterations were not perceived by.highway planners and designers as important, if they were perceived at all. 13

PAGE 29

Throughout the 1960's impacts to natural systems from highways and a myriad of other sources continued to grow. In response, the environmental movement in America was hom and Congress was pressured into passing legislation to protect the environment. I review statutes that pertain to highways in detail in the next section of this chapter. However, even with these statutes in place, the attitudes of highway professionals and the design standards required to receive highway funding continued to make it difficult to design highways to anything but the needs of the roadway itself. Although other considerations were not forbidden, safety and efficiency at the posted speed continued to be the only criteria recognized in AASHTO's design standards. Because the vast majority of highway professionals did not share the public's new-found environmental awareness (Mowbray, 1969; Lewis, 1997) and because of their deeply entrenched negative attitudes about the input from narrow interpretation of these standards continued. Most highway projects through the 1970's and 1980's were designed to the roadway's needs and then minimally modified to accommodate other regulatory requirements (Wick, 1995). Traditional Conservation Planning and Highways As reviewed below, conservation planning in the u:s. has a long history of ignoring how conservation goals might be attained in context of the built environment. While this outlook certainly includes a lack of consideration about how highways impede conservation goals, as well as a lack of interest in how highway planning and design might by modified to minimize these impacts, it would be incorrect to claim that highways have been singled out for this attitude. Two strong themes in Western culture, the tendency to see nature as something separate and alien from humans, and a tendency to place nature on a pedestal have 14

PAGE 30

created a strong schism between planning for conservation and planning for all aspects of the built environment. Actions that can be construed as wildlife conservation planning began in the American Colonies shortly after settlement with the passage of laws specifying the hunting season for deer in Rhode Island, and continued during the 1700's and 1800's with the passage of similar statutes throughout the country (Leopold, 1933; Andrews, 1999). During this entire period, wildlife was treated separately from the management of other natural resources such lumber and water, and there was no attempt to account for wildlife values as agricultural clearing, settlement, and road building occurred. However, by the mid-1880's it was becoming apparent that animal populations were not unlimited, and that additional management would be required to prevent their eventual destruction (USDI, 2001). The advent of recreational (as opposed to subsistence) hunting and wildlife observation as a hobby also began in the late 1800's, and this public interest in multiple uses for wildlife required a different approach to management as well. At the tum of the 20th century, President Roosevelt and Forest Service head Gifford Pinchot began to champion an approach to natural resource management that they termed "conservation through wise use." This movement sprang from a belief that resources needed to be conserved in order to provide future benefits (raw materials, food, amusement) to humans, but it represents the ftrst widespread linking of the term "conservation" to the stewardship of natural resources. The movement also had two other important principles, the recognition that all natural resources comprised an integrated whole, and the recognition of science as an important tool for informing natural resource management. This interest in the scientification of natural resource management, combined with the newly applied idea of 15

PAGE 31

conservation, helped to create the field of resource management. This new class of experts was also, in effect, America's first professional conservation planners. However, although these experts increasingly thought of nature as a whole, they did not break from the traditional Western cultural view that nature and humans are separate (Western, 1989). This view was expressed in creation of the first National Parks in the late 1800's and the first wildlife refuges in the early 1900's, which focused the practice of wildlife conservation on setting patches of habitat aside from humans. Despite the vast growth in the knowledge and theory that supports the practice of conservation planning, discussed below, this mind-set remained embedded in conservation planning throughout the 20th centwy due to both philosophical outlooks and for practical reasons. The concept of delineating and managing a separate patch for wildlife fits in well with both western cultural norms and with normal science's orientation towards simple, orderly solutions based on single, knowable truths (Lister, 1998). Practically, planning and managing a wildlife refuge, a single-use patch with definite boundaries, is a simple task, as compared to the ''wicked" problem (Rittle and Webber, 1973) of planning and managing an ill-defmed area for multiple uses. Concurrent with the adoption of a scientific approach to planning for and managing wildlife was the growth of the science of ecology, which provided an additional source of expert information for natural resource mangers. The Ecological Society of America was officially constituted in 1915, and the study of ecology flourished from the 1920's onward. As the field grew, an understanding of humans' place in nature and the profound effects ofhuman actions on natural systems emerged. Popular works by Aldo Leopold (1949) and Rachel Carson (1951, 1955, 1962) made these new ecological concepts accessible to non-scientists. However, the widespread recognition of humans as an integral part of the natural world did not 16

PAGE 32

translate into an integration of the management of human infrastructure and natural resources. Although the birth of the environmental movement in the 1960's began to break down the long held societal nonn that nature was separate and alien form human endeavor, it was replaced to some extent with the notion that humans were so damaging to nature that it had to be kept separate in order to survive. Thus, no matter to what school of thought wildlife conservation planners subscribed, there was an underlying tendency to ignore the built environment that increasingly penetrated and surrounded wildlife habitats. This outlook was probably compounded by their professional training, which quite naturally focused them on their own field of expertise. The ecological subdiscipline of reserve site selection offers a clear illustration of this single-mindedness. This area of theory, regarding the best size and arrangement of reserves, as well as the best schemes for connecting them (e.g., Diamond, 1975; Pickett and Thompson, 1978; Soule and Simberloff, 1986; Simberloff and Cox, 1987) was developed by ecologists as a direct result of their interest in conservation planning. It was built largely upon the theory of island biogeography (MacArthur and Wilson, 1967), which had been in tum developed from basic ecological principles, that by defmition are not concerned with influences of the human environnient. Therefore, it is not surprising that most reserve site theories focused strictly on natural constraints, and did not account for the human land-uses and infrastructure that might already be in place. Additionally, little interest was accorded as to how areas surrounding reserves, unprotected and subject to multiple uses including highway construction, might be planned for, designed, or managed to increase its value to wildlife (Prendergast et al., 1999). During the 1980's, the reserve configuration discussion and its interplay with attendant topics such as genetics and population biology helped give rise to the 17

PAGE 33

field of conservation biology (Soule, 1985). Its practitioners dominate the field of conservation planning today. Although this latest class of conservation planners does not unerringly heed its roots, conservation biologists continue to frame much of their effort in terms of reserve creation (e.g., Noss, 1992; Soule and Terborgh, 1999; Hoctor et al., 2000; Margules and Pressey, 2000). Although most authors acknowledge that meeting conservation goals may be difficult due to competing land uses, only rarely do they examine the constraints this reality places on a reserve-centric approach to conservation. Likewise, there has been little emphasis on developing approaches that focus on meeting conservation goals in areas that have not been set aside as reserves (Prendergast et al., 1999). Among wildlife experts, the concept of conservation planning remains largely a biologically based process, conducted independently of planning for other activities and projects that also affect land use. Despite the narrow focus of many conservation planners, fmding ways to create and maintain habitat that can sustain wildlife while subject to other uses has received at least theoretical attention in other fields, notably land-use planning and landscape ecology (e.g., McHarg, 1969; Western, 1989; Soule, 1991; Hansen et al., 1991; Hansen et al., 1993; Shafer, 1994; Dale et al., 2000). Never the less, efforts to integrate wildlife considerations into planning for the built environment are by no means standard procedure. However, some important first steps to integrate conservation planning into highway planning and design have been taken and are discussed below. The Current Institutional Setting for Reducing Wildlife/Highway Conflicts The problem of highway-caused impacts to the natural environment is becoming more widely acknowledged by highway professionals. For example, the 18

PAGE 34

Transportation Research Board established the Committee for Environmental Analysis in Transportation in 1991, and sponsored workshops at its annual conference in both 2001 and 2002 to defme the concept of"environmental stewardship" and create a framework for state transportation departments to institute it. AASHTO created a Standing Committee on the Environment in the early 1990's and highlighted environmental stewardship as one of the three focal topics at its 2002 annual meeting. Likewise, there is an increasing recognition of highway-related issues in the conservation community. Academics, agency personnel, and environmental groups are studying and discussing conflicts between highways and the natural environment. Examples of the growing interest in the topic include the special section on the ecological effects of roads in issue 1, volume 14 (2000) of Conservation Biology, Richard Forman's new text Road Ecology (2002), and the 2001 International Conference on Ecology and Transportation which was attended by over 340 individuals and sponsored by 11 organizations, including the U.S. Forest Service, U.S. Fish and Wildlife Service, the Wetlands Division of the Environmental Protection Agency, Defenders of Wildlife and the Humane Society of the United States. There is no single reason that the negative impacts of highways and wildlife have lately become a topic of concern. An important general factor is probably the increasing awareness among many sectors of American society that we live in a world of finite resources. Two specific manifestations ofthis cultural change playing a role in raising the profile of highway/wildlife conflicts are: 1) the regulatory mandates that currently control federal funding of highway projects and 2) the advent of the ecosystem approach in natural resource management. Both of these themes are shaping the current institutional setting that is encouraging 19

PAGE 35

conservations professionals to press their concerns about highways and highway professionals to address wildlife/highway conflicts as part of project planning and design. The Regulatory Nexus to Consider Wildlife in Highway Planning In the mid Congress began passing legislation to force highway departments to plan and design highways that would ameliorate negative impacts. Regulations directly applicable to wildlife issues included the Department of Transportation Act of 1966, which forbade the use of federal dollars on projects that would adversely impact publicly owned land that was used as a park, recreation area, or wildlife refuge, unless there is "no feasible and prudent alternative to the use of such land" [23 U.S.C., 138 section ( 4)(f)], and the 1970 Federal-Aid Highway Act which set noise standards, required project planners to coordinate with state.and local quality to meet standards forth by the Clean Air Act of 1970, and mandated that the economic, social, and environmental effects of a highway project be considered before it could go forward. During the 1960's and early 1970's, Congress also passed a number of laws specifically regulating impacts to the environment, which in tum could be applied to reduce highway impacts. The regulation with the most direct application to wildlife is the Endangered Species Act (ESA). However, statutes that regulate land use can also effectively prevent impacts to. wildlife. Depending on the location of a highway project, a variety of environmental regulations can influence its placement and design, to the benefit of local wildlife. I list statutes that potentially have an extensive effect in Table 1.1. Each ofthese acts requires that adverse impacts to the resources they address be avoided or minimized. Conscientious application of these 20

PAGE 36

regulations can have a significant effect on the level of impact created by a highway project. However, generic environmental regulations do not require highway specific mitigation, and highway planners and designers have historically tended to ignore or only minimally complied with their mandates. Without a genuine commitment to the intent of these laws among highway experts, the success of these statutes often depends on the amount of external pressme a regulation's champion applies on the highway project planning process. As a result of the uneven application of these statutes, public dissatisfaction with the process of highway planning and design as well as the outcomes ofhighway projects continued throughout the 1980's. In response, Congress strengthened the basis for minimizing the impacts of highways to the environment through additional legislation. The lntermodal Surface Transportation Efficiency Act (ISTEA}, passed in 1991, was designed mainly to reduce transportation's impact to the human environment, but applied equally to the natural environment. In particular, IS TEA explicitly repealed the use ofthe AASHTO standards as a prerequisite for project funding and required that 10 percent of funds allocated for surface transportation be set aside for non-pavement transportation enhancement (TE) projects. Language within the Act emphasized "preseniing and protecting environmental and cultural values affected by transportation facilities" (FHWA 1997, p. v). This created the first legal mandate for reducing environmental impact through sensitive project planning and design. 21

PAGE 37

Table 1.1 Environmental statutes with potentially extensive effects on the placement and design of highways in the U.S. Regulation Name Wilderness Act Highway Beautification Act Wild and Scenic Rivers Act National Environmental Policy Act Clean Air Act, Transportation Confonnity Rule Clean Water Act Costal Zone Management Act Endangered Species Act Coastal Barrier Resources Act Year Passed 1964 1965 1968 1969 1970 1972 1972 1973 1982 Influence on Highways Forbids any type of development on lands designated as wilderness by congress. Controls outdoor advertising and placement of junkyards adjacent to interstate and primary highway systems. Provides funding for roadside landscaping. Forbids certain impacts to rivers designated as wild and scenic by congress. Requires disclosure of environmental impacts caused by all federally funded actions, including highway projects. Requires that transportation projects conform to State air quality implementation plans. Requires the restoration and preservation of chemical, physical, and biological integrity of the nation's waters through prevention, reduction and elimination of pollution from all sources. Forbids impacts to jurisdictional wetlands. Preserve, protect, restore, and enhance resources of the coastal mne. Forbids negative impacts as a result of any federally funded project to fish, wildlife, and plant species designated as threatened or endangered. Minimize damage to fish, wildlife, and other natural resources due to projects occurring within the boundaries of a designated coastal barrier unit During the mid and late 1990's, additional highway specific legislation expanded the vision ofiSTEA and specifically emphasized reducing impacts to wildlife. The 1995 National Highway System Designation Act (NHS) specifically states that highway design may be tailored to the natural environment and to reducing environmental impacts. Although the statute is permissive rather than mandatory, it provides additional legal support for making the environment a primary consideration in highway design. Then, in 1998, the Transportation Equity Act for the 21st Century (TEA-21) expanded the list ofTE activities established by ISTEA to include projects ''to reduce vehicle-caused wildlife mortality while" maintaining habitat connectivity" [2J U.S.C., section 101(a)(35)]. 22

PAGE 38

Another important development helped to create a regulatory nexus for reducing highway impacts to wildlife was the FHW A's adoption of the planning model known as Context-Sensitive Design (CSD) in 1997. Responding to the mandates ofNHS and TEA-21, the FHWA officially approved CSD as a tool to make considerations of both human and the natural environment a focal point in the design of highway projects. Although this guidance to the states was not legally binding, the policy document Flexibility in Highway Design legitimatized the concept of allowing a roadway's surroundings to guide its form. This document, written as a companion volume to AASHTO's book of geometric design standards advocated (FHW A, 2001 ): Projects that satisfy a full range of stakeholders Stakeholder agreements forged early in the project and amended as needed Projects that provide safety for both the user and the community Projects that are in harmony with the environment Projects that are built with minimal disruption to the environment Projects that are seen as having lasting value to the environment Highways and the Ecosystem Approach to Conservation Planning The growing interest among conservation planners in addressing wildlife/highway conflicts is in part a spillover effect from the legal affecting highway planning. However, an ongoing paradigm shift ip the practice of wildlife management also plays a strong role. Wildlife conservation planning in the United States traditionally focused on creating.and protecting reserve areas. When the human population was small, the surrounding matrix of unprotected area was largely undisturbed and augmented America's relatively small total reserved area. However, as human populations increase, they place greater pressures on this 23

PAGE 39

matrix area, and its ability to provide conservation values decreases, threatening the function of many reserves. As a wildlife professionals are recognizing that simply creating preserves and managing them as natural "islands," separate from the sea of development that is occurring around them, is unlikely to achieve conservation goals. Instead, these professionals realize they must explicitly acknowledge and incorporate the built envrronment, such as highways, into their planning to secure success. The spread of this new attitude is illustrated by the advent of ecosystem approaches for natural resources planning and management, beginning in the early 1990's. The ecosystem approach is generally acknowledged as being a concept, rather than a prescriptive methodology for achieving a particular goal (Grumbine, 1994; Born and Sonzogni, 1995; Brossard et al., 1998; Yaffee 1998). The practical application of an ecosystem approach varies, and it may be referred to as integrated environmental management, watershed management, or ecosystem management, among other similar names. The common threads of planning and management practices that fall under the ecosystem rubric are reliance on data about the system of interest, utilization of adaptive management, promotion of collaboration between different disciplines, agencies, and stakeholder groups, and acknowledgment of the role ofvalues as well as science when setting and implementing conservation goals. Additionally, ecosystem approaches acknowledge ecosystem complexity and dynamism, multiple temporal and spatial and ecological, as opposed to political, boundaries (Slocombe, 1993; Grumbine, 1993; Born and Sonzogni, 1995; Christensen et al., 1996; Haeuber, 1996; Brossard et al, 1998; Yaffee, 1999). 24

PAGE 40

Examples of the paradigm shift in the conservation planning community include adoption of an ecosystem approach by the federal government during the 1990s. The U.S Forest Service approved "ecosystem management" as the basis for natural resource management in 1992. A similar approach was.subsequently institutionalized throughout the federal government in a Memorandum of Understanding (1995) that was signed by 14 federal agencies. The Nature Conservancy (TNC) provides another example. Historically, TNC was dedicated to simply purchasing and preserving reserve areas, but expanded its mission in 1996 to include management of landscape-level processes through an ecosystem approach (Poiani, 1996). Two themes from the ecosystem approach to conservation planning encourage resource managers to consider the effects of the built environment, including wildlife/highway conflicts. The emphasis on recognizing ecological boundaries instead of the artificial boundaries of management areas promotes conservation professionals to acknowledge the impacts of infrastructure such as highways on both the protected areas and wildlife populations that move in and out of those areas. The emphasis on interdisciplinary collaboration sets the stage for interactions with highway planners and designers in order to fmd collaborative solutions that can reduce impacts. The Current Practice of Reducing Wildlife Highway Conflicts As discussed above, the conditions under which highway professionals plan and design highway projects have changed substantially since the 1960's, and especially during the past decade. Additionally, changing paradigms in natural resource management mirror this growing mandate for highway professions to include non-pavement considerations in their work. However, the actual practice of 25

PAGE 41

reducing wildlife/highway conflicts remains in its infancy, and the procedures for incorporating mitigation into highway projects are haphazard. Most decision making related to this type of conflict mitigation relies on expert opinion rather than data, and there are no objective standards in either the highway or the conservation community to determine if mitigation is sufficient. Efficient implementation of projects designed to reduce conflicts is impeded by a lack of information about the nature ofwildlifelhighway interactions. Despite the relatively extensive literature documenting impacts of roads on the environment (reviewed above and in: Bennett, 1991; Spellerberg, 1998; Trombulak and Frissell, 2000), the ecological effects of roads have yet to receive attention as a unified category of research or conservation planning by ecologists and conservation biologists (Hourdequin, 2000). In particular, there is especially little data about how wild animals interact with the roadway as well as where these interactions are most likely to occur. This lack of good information affects the ability of highway professionals to confidently plan mitigation projects. Roadway designers can reduce wildlife/highway conflicts by modifying the highway itself, for example by including wildlife underpasses, or they might choose alignments that avoid sensitive wildlife areas all together. However, because these types of design modifications are costly, the amount of mitigation that can be included may be limited by theproject budget. Deciding how to allocate limited funding to do the most good is difficult without sound information. The willingness of wildlife professionals to participate in wildlife/highway conflict mitigation planning is also impeded by the lack of information about wildlife/highway interactions. A hallmark of the ecosystem approach to natural 26

PAGE 42

resource management is the reliance on data to create management strategies for the system of interest. Conservation planners increasingly demand good information in order to formulate their own plans or to buy into conservation oriented strategies designed by others. The unwillingness of some wildlife professionals, especially the regulatory personnel, to endorse mitigation projects proposed by highway professionals, is a strong disincentive for highway designers to include wildlife considerations in their plans. Summary Negative impacts of highways on wildlife include highway-induced habitat loss and alterations, habitat fragmentation, direct mortality, and disturbance. Although the list of documented impacts is substantial, less is known about the mechanics of wildlife/highway interactions. This type of basic information is needed to mitigate existing impacts and avoid future impacts. As the historical barriers between highway and conservation planners break down due to legal mandates and the advent of the ecosystem approach to conservation planning, the need for sound information to inform decision-making is growing. 27

PAGE 43

CHAPTER2 A REVIEW OF RESEARCH RELATED TO IDENTIFYING WILDLIFE HIGHWAY CROSSING ZONES Introduction Most of the literature documenting interactions of wildlife and roads (reviewed in: Bennett 1991, Spellerberg 1998, Trombulak arid Frissell2000) focuses on simply quantifying the negative effects of roads on wildlife and their habitats. To date, researchers have spent relatively little energy investigating the variables that regulate how wild animals interact with roads and that might be used to predict the location and severity of negative impacts. In particular, few studies have attempted to determine if there are specific locations where animals are more likely to cross highways, and to identify the characteristics associated with these locations. I review the small amount of existing literature that pertains directly to these two questions in this chapter. Additionally, I examine literature regarding roadside habitat use, the use of crossing structures by wildlife, and the locations of animal/vehicle collisions (A VC). Because these types of studies also investigate where wild animals interact with roadways, they provide useful insight into the variables that are likely to be associated with the locations where animals cross the roadway. The research that I review is summarized in Table 2.1, then examined in detail below. I begin by covering the three attendant topics, and close by summarizing the existing research 28

PAGE 44

Table 2.1 A summary of variables demonstrated to be important in determining the locations where animals interact with roads and highways Author Research Important Variables Summary of Results Alexander and Waters, 2000 Cross Topography Crossing associated with flat slopes, south-to-west exposures, low topographic complexity Allen and McCullough, 1976 AVC Food A VCs occurred near cover types providing food Bashore et al., 1985 AVC Food A VCs occurred near cover types providing food Bellis and Graves, 1971 AVC Food A VCs occurred near cover types providing food Carbaugh et al., 1975 Cross, Hab Roadside characteristics, Food Roadside habitat uses associated with cover types providing food, crossings avoided roadside barriers, steep cuts & fills Clevenger et al., 2001 Struc Roadway characteristics For most species structure use increased with traffic volwne Clevenger and Waltho, 2000 Struc Human activity Some species only use structures with low levels of activity Feldhamer et al., 1986 Hab Food, A VCs occurred near cover types providing food Finder et al., 1999 AVC Cover, Topography A VCs occurred near forest cover, drainages Gibeau et al., 2002 Hab Habitat, Human activity Roadside use associated with preferred habit, low levels of human activity N Haas, 2000 Cross Cover; Habitat fragmentation Structure use associated with unfragmented habitats and cover 10 near the entrance Hubbard et al., 2000 AVC Topography, Habitat AVCs associated with large cover-type blocks, drainages fragmentation lobar and Mayer, 1999 AVC Habitat Roadside use associated with preferred habit Lovallo and Anderson, 1996 Hab Habitat Roadside use associated with preferred habit Lyon, 1979 Hab Cover Roadside use associated with forest cover types Mamalis, 1995 Cross Roadside characteristics Crossing avoided roadside barriers Puglisi et al., 1974 AVC Food A VCs occurred near cover types providing food Reilly and Green, 1974 AVC Food A VCs occurred near cover types providing food Rodriguez et al., 1996 Struc Cover Structure use associated with cover near the entrance Romin and Bissonette, 1996 AVC Topography, Cover A VCs associated with drainages, non agricultural cover types Rost and Baily, 1979 Hab Cover Roadside use associated with forest cover types Singleton and Lehmkuhl, 2000 Cross Habitat fragmentation, Human Roadside use associated with unfragmented habitats, low levels activity of human activity Yanes et al., 1995 Struc Cover Structure use associated with cover near the entrance ---------------------------... -

PAGE 45

directly concerned with identifying actual crossing-locations and quantifying their characteristics. Roadside Habitat Use A widely documented effect of roads on wildlife is the avoidance of areas adjacent to the roadway (Bennett 1991, Spellerberg 1998, Trombulak and Frissel12000). However, some studies indicate that certain habitat characteristics appear to mediate this response. It is important to consider the results of studies that document this effect. Areas adjacent to highways that are more likely to be used despite the presence of the roadway may also be the locations where animals are most likely to cross the road. The quality of roadside habitat in relationship to a species' preferred habitat type is one factor that appears to play a role in an animal's willingness to approach roads. Grizzly bears in Alberta, Canada (Gibeau et al., 2002) and bobcats in Wisconsin (Lovallo and Anderson 1996) were both more likely to use areas near roads when these roads were located in preferred habitat types. The response of grizzly bears to preferred habitats was weaker along roads with high traffic volwnes (Gibeau 2002). Lovallo and Anderson (1996) did not describe the characteristics of roads in their study, nor did they separate them by type in their analysis. Food and the presence of cover in the form of trees are also two habitat factors that influence roadside habitat use. Mule deer in Colorado avoided all roads, including highways, under all conditions, but showed a reduced avoidance response in pine forests and juniper woodhmds, as compared to shrub habitats (Rost and Bailey 1979). Elk in Montana showed a similar pattern, increasing their use of roadside habitats as canopy cover increased (Lyon 1979). Roadside habitat use by white-tailed deer (Odocoileus 30

PAGE 46

virginianus) in Pennsylvania was most influenced by the presence of food. In locations where the roadside habitat was primarily wooded, deer often entered the right-of-way to graze on grasses growing along the shoulder and in the median. In locations where the roadside habitat was primarily open, deer foraged away from the right-of-way (Carbaugh et al., 1975, Feldhamer et al., 1986). Animal/Vehicle Collision Locations Along Highways Although accounts detailing the phenomenon ofroadkill in the U.S. were published almost as soon a8 widespread use of the automobile began (e.g., Stoner 1925, Cottam 1931 ), researchers did not undertake formal studies about A VCs until the early 1970's. These initial studies focused on white-tailed deer and were conducted in Pennsylvania (Bellis and Graves 1971, Puglisi et al., 1974) and Michigan (Reilly and Green 1974, Allen and McCullough, 1976.) Because the focus of these studies was on the temporal rather than the spatial distribution of collisions, the information they contain regarding A VC locations is qualitative rather then quantitative. However, the authors of all' four studies concur that high kill areas were associated with locations that had cover types that attracted foraging deer, as well as high traffic volumes. Investigations conducted since the 1980 have utilized more rigorous approaches to quantify the features associated with AVC locations. Using logistic regression, Bashore et al. (1985) determined that variables representing longer driver line-of sight distances and a mix of cover types played the primary role in identifying higher density deer roadkill zones on Pennsylvania highways. They hypothesized that long sight-lines encourage motorists to drive faster, making it difficult for them to avoid deer on the roadway, and that mixed cover types attract more foraging deer because they provided both food and cover. 31

PAGE 47

The most recent studies used GIS to analyze the characteristics of A VC locations. Finder et al. (1999) used remotely sensed site and landscape data to determine that forest cover was the variable most strongly associated with high kill zones in Illinois. In addition, gullies adjacent to the road, riparian zones, and public recreational land within 0.8 km of a point further increased the probability that it would be classified as "high kill." Hubbard et al. (2000) analyzed deer/vehicle collision locations in Iowa using a similar approach, and determined that collisions increased as cover type patch size, number of bridges per 1.61 km highway segment, and lanes of traffic increased. Although the locations of AVC provide an indication ofwhere wild animals interact with highways, they may be a poor surrogate for determining where animals prefer to cross highways. AnimaVvehicle collision data record only unsuccessful crossings attempts, ai:td only a small subset of those. A VCs are generally reported only when property damage or injury to the vehicle's occupants warrants a call to the police. Based on estimates from state transportation and wildlife management agency field personnel, Romin and Bissonette (1996a) reported that nationwide, orily 17 to 50 percent of A VCs that occur are reported. Additionally, because responding officers usually estimate the location of the AVCs to the nearest milepost, only the general location of wildlife highway crossing zones can be gleaned from this data. Thus, the locations of actual road-killed wildlife observed in the field, instead of reported AVC, may be a more accurate indicator ofthe locations where wildlife interact with highways. Two published studies address the factors associated with lethal interactions, as indicated by carcasses. The distribution of road-killed mule deer in Utah was positively correlated to locations where large drainages bisected the road and to areas with non-agricultural cover types (Romin and Bissonette 32

PAGE 48

1996b ). lnbar and Mayer (1999) used multiple regression and correlation analyses to determine that high mortality areas for armadillos (Dasypus novemcinctus) in Florida were associated with preferred habitat types in winter and areas with high night-time traffic volumes in summer. Use ofHighway-Crossing Structures by Wildlife Wild animals that cross highways can do so either at-grade, by walking across the hard top, or they may utilize structures to cross over or under the road's surface. The bulk of the existing research pertaining to the locations where wildlife cross highways is focused on animals' use of such structures. This research looks at use of structures constructed specifically for wildlife as well as at opportunistic use of structures installed for other reasons. Studies have been conducted along both fenced (Foster and Humphrey1996, Roof and Woodling 1996, Clevenger and Waltho 2000) and unfenced highways (Hunt et al., 1987, Yanes et al., 1995, Rodriguez et al., 1996a, Hewitt et al., 1998, Haas 2000, Clevenger et al., 2001). In aggregate, these studies indicate that animals use a wide variety of structures, including bridges, culverts, and tunnels, as well as the structures built specifically for wildlife, to cross under highways and high-speed rail corridors. However, studies focused on underpass use, along either fenced or unfenced highways, cannot indicate anything about where animals prefer to cross transportation corridors at-grade. Studies conducted along highways fenced to prevent crossing except at s!Illctures can only test the preference for different structures and structure location. Wildlife using structures along unfenced highways and railways can cross at any location, but in all the existing studies, data were only at the structures. Therefore, conclusions can still only be made about structure preference. 33

PAGE 49

Research about overpass use by wildlife is very limited and suffers from the same limitation as the underpass research. Wildlife-friendly overpasses have been built in only two locations in North America (Banff National Park, Alberta, Canada, and 115 in Utah, USA) and research regarding use of these structures is unavailable. Overpasses have been more widely constructed in Europe, but the single available English-language report summarizing their use indicates that only the structures were monitored (Pfister 1997). Thus, like the underpass research, no conclusions can be drawn from it about where animals prefer to cross the highways in question at-grade. Never-the-less, the results of these studies do provide some insight into the habitat variables associated with structures most frequently used, and it is possible to infer that some of these same variables may influence the choice of at-grade crossing locations as well. Results of research that quantitatively compared crossing rates and the surrounding habitat at multiple structures indicate that habitat elements from both the localand the landscape-scale play a role in determining a structure's rate of use. At the landscape-scale, Haas (2000) found that species sensitive to humaninduced habitat fragmentation use underpasses associated with such habitats less than the underpasses associated with more intact habitats. At the local-scale, researchers correlated a variety of habitat features to underpass use. The level of human activity is a significant factor for some speCies. Predators, but not ungulates, tend to avoid structures with high levels ofhuman activity (Rodriguez et al., 1996a, Clevenger and Waltho 2000). The presence of cover at or near the entrance to an underpass increases the rate of use by many smalland mid-sized species, including mustelids, canids, felids and 34

PAGE 50

lagomorphs (Yanes et al., 1995, Rodriguez et al., 1996a, Haas 2000, Clevenger et al., 2001). The characteristics of the roadway itself, such as pavement width, traffic volume, and associated noise levels are also important to these smalland mid sized species (Yanes et al., 1996, Clevenger et al., 2001). Response to these variables varied by species and the authors hypothesized that this was due to each species' vulnerability to predation while crossing the clear zone associated with the roadway. These local-scale habitat factors appear to override structural characteristics of the crossing structure itself in some cases, especially for predator species (Rodriguez et al., 1996a, Yanes et al., 1996, Clevenger and Waltho 2000). Locations.of At-Grade Highway Crossings by Wildlife Studying at-grade animal movements along an entire highway corridor, as opposed . to concentrating on focal points such as underpasses, is difficult. Methods for recording corridor-wide data include direct observation of animals crossing the highway, continuous monitoring of radio-collared animals, and using tracking techniques to infer animal moyements. Direct observation is not a very efficient approach, as it is labor intensive and subject to bias animals will_ only be seen crossing roads in location that are under observation during that particular event. However, while conducting a study of roadside habitat use by white-tailed deer, Carbaugh et al. (1975) opportunistically observed 160 deer in the act of crossing the eastbound lanes of I -80 in central Pennsylvania. The authors did not provide a detailed description of the highway (lane width, median width, etc.), and did not analyze the crossing data quantitatively. Nevertheless, they provide a valuable qualitative description of crossing behavior. Observed crossings were focused on areas that had not been modified during highway construction by cutting and filling. Hill-valley intergrades 35

PAGE 51

and areas that sloped upward from the highway were used most often. The authors hypothesized that deer were guided to these unmodified areas by cuts, fills and guardrails. Deer rarely jumped guardrails to enter the roadway, although they jumped them readily to exit it. Continuous monitoring of radio-collared animals provides another option for recording highway their crossing locations. This approach is also labor-intensive when using standard VHF collars. Consequently, the staffmg constraints of most wildlife studies would require choosing a single focal animal to follow during each monitoring session, resulting in bias and limited data sets. Therefore, this method apparently has not been widely applied. A study of highway crossing behavior by wolves in Wisconsin (Kohn et al., 1997) that used this technique did not get an adequate sample size of crossing events for analysis. The advent of radio collars that can continuously relay a GPS derived location to a receiver via satellite is making continuous monitoring of an animal's movements more practical. However, the application of this technique is currently limited, as these transmitting devises require significant battery power. Because large batteries are heavy, these types of collars can only be placed on large animals, such as bears and bison. A study of grizzly bear movements in relationship to roads in Montana is in progress (Waller and Servheen 1999), but results are not yet available. Roadside tracking techniques provide a more versatile option for recording animal movements across roads and provide a reasonable trade-off between labor required and data acquired. However, a naturally present, suitable roadside tracking medium is needed throughout the corridor under study to record animal movements, because creating a medium in such large areas is infeasible. Reliable snow cover during the winter months provides perhaps the best opportunity to conduct this type ofresearch, and three studies, (Mamalis 1995, Singleton and Lehmkuhl2000, 36

PAGE 52

Alexari.der and Waters 2000) have taken advantage of this opportunity. Mamalis (1995) recorded crossing behavior of all species of wildlife present along the Trans Canada Highway in Banff National Park, Alberta, during the January-March period of 1995. The distribution of crossings was significantly different from random ("J! =15.51, 8 d.f., p<0.05) and crossings were more common than expected in several locations. Snow depth, the only factor that was quantitatively assessed, did not significantly influence choice of crossing location. Qualitative descriptions of crossing zones indicated that existing wildlife travel corridors and roadside barriers might have influenced where animals crossed the highway. Singleton and Lehmkuhl (2000) recorded crossing location of all species present from January through march 1999 and from December 1999 through March 2000 along a 30 mile stretch ofl-90 between Cle Elum and Snoqualmie Pass in Washington, USA. They mapped their results and visually identified distinct clusters of crossings, but did not quantitatively analyze the extent of the clusters or the habitat features associated with them. Qualitative descriptions of these locations indicate that they were generally associated with intact natural cover types, low levels of human disturbance, and landscape linkage zones that the authors had . . previously identified through a GIS-based analysis. Alexander and Waters (2000) recorded crossing behavior of all species of wildlife present along three 30 km sections ofhighway in BanffNational Park, Alberta, Canada during the winters of 1997/1998 and 1998/1999. They identified high frequency crossing zones ranging from 250 to 2000 m in diameter. The authors then used chi-square tests to compare the characteristics of crossing locations to expected values. Results indicated that general predictors of movement for all species included slopes facing south, southwest and west, slopes angles of 5 or less, and areas oflow topographic complexity. However, the authors did not 37

PAGE 53

indicate at what scale these measurements were made, i.e., they did not state how far away from the road's edge their measurements of the surrounding habitat extended. Summary Reviews of literature pertaining directly to where wild animals choose to cross roads indicate that at-grade crossings do not occur at random locations. This conclusion is further supported by the non-random distribution of AVCs and roadkilled carcasses, both of which can be viewed as indicators of crossing behavior, and by the preferential use of crossing structures associated with certain habitat features. Additionally, non-random variation in the intensity of roadside habitat use also suggests that wild animals are more likely to approach highways in certain locations. The literature pertaining to each of these topics indicates that a variety of factors influence how and where wild animals interact with the roadside. The most important variables identified in this literature review were the presence of food, cover, preferred habitat, drainages that intersect the roadway, and characteristics of the roadway itself. Animals also avoided roads in areas with high levels of habitat fragmentation and, for many species, high levels of human activity. Although only one of these factors, habitat fragmentation, was explicitly investigated in terms of scale, it is also clear that variables from both the local and the landscape scale play a role in the location of wildlife/highway interactions. However, because of the disparate study designs used in the-reviewed literature, it is not possible to compare the influence of localversus landscape-scale variables, or to rank these variables in order of importance. 38

PAGE 54

Overview CHAPTER3 IDENTIFYING WILDLIFE HIGHWAY -CROSSING ZONES IN THE COLORADO ROCKIES: RESEARCH BACKGROUND Introduction The goal of my research was two-fold. First, I wanted to determine if the locations where wild animals cross highways are correlated. with defmable characteristics of the surrounding environment and the roadway itself. Then, I was interested in how this type of information could inform the highway design process to reduce wildlife/highway conflicts. I hypothesized that: Wild animals do not cross highways at random and the locations that they use can be correlated with features of the roadside and the surrounding habitat. The identification of likely crossing locations is an accessible process that can be incorporated into highway design to reduce wildlife/highway conflicts. In this chapter I describe the rationale for and methodology of the research that I conducted to examine the first part of my hypothesis. I begin by reviewing my approach, then justify the comparison variables that I selected, and describe my study sites. The remainder of the chapter is dedicated to describing my methods of data collection and analysis in detail. The data collection descriptions are arranged 39

PAGE 55

according to the type of data that was collected. Explanations of the data analysis are arranged according to the question being examined. The arrangement of the analysis descriptions serves also to organize Chapter 4 (results) and Chapter 5 (discussion of the results). The second part of my hypothesis is examined in Chapter 6. The Research Approach My research identified locations where animals crossed unfenced highways at grade. I then compared the crossing locations to random locations along the highway, to determine if and how these two types of locations differed. I defined highways as paved roads, at least two lanes in width, and with posted speeds of 55 kmlh or greater. Other types of roads (e.g., unpaved, single lane, and/or low speed) have a variety of effects on wildlife, including behavior modification, introduction of competitors, and increased human activity (Trombulak and Frissell 2000), but are apparently crossed by wildlife (C. Apps, pers com 1998; McKelvey et al., 1999). Therefore, I did not consider these types of roads. Features that may differ between crossing zones (CZs) and random locations could come from either the local or the landscapescale. "Local" and "landscape" are relative terms that must be defined by the context of their application. For this study, I considered the local scale to be the characteristics of the roadway itself as well as topographic and vegetation features within 100 m of the roadway. I defmed the landscape scale as the general landforms and cover types encompassed by the ridge lines that provided visual boundaries of the area surrounding the highways in my study areas. 40

PAGE 56

To identify CZs for comparison with random roadside locations, I recorded animal activity at the roadside along two highway corridors, one with a two-lane highway and one witha four-lane highway. I chose two highways with different footprints because I wanted to know if highway design affected the highway-crossing locations of wildlife. I analyzed animal roadside activity data to identify the areas that wild animals crossing the roadway used most often. I monitored roadside wildlife activity along the entire length of both highway corridors year-round, as opposed to seasonally and/or only at focal points such as underpasses. Additionally, I monitored bridges and over-sized culverts to determine if animals were actually using them cross under the highway, and whether the presence of an underpass inflenced at-grade crossing rates and locations. Relevance of the Measured Variables I chose the variables to compare crossing with random locations by considering the existing research pertaining to wildlife/highway interactions. In addition, I considered the suite of roadside habitat characteristics whose form, and presence or absence of which are under the direct control of highway designers and builders. A complete list of all the variables that I measured is given in Table 3.1, and the rationale for their inclusion is discussed below. Although I chose to categorize variables as landscape-scale or local-scale (Table 3.1) when describing study methods and results, it is important to note that they could be divided in other ways. Other classification schemes include habitat characteristics versus highway characteristics and measurements that must be made in the field versus measurements that can be made "remotely" from digital data layers or aerial photographs. Habitat characteristics are a mix of localand landscape-scale variables, where as highway characteristics are, by the defmition of 41

PAGE 57

my stlidy, local scale only. The measurements that I made in the field for this study were also all at a local-scale, but the "remote" measurements encompassed both localand all landscape-scale measurements. Table 3.1 Variables measured to compare crossing-zone (CZ) locations to random locations Local or Habitat or Field or Variable Description Landscape Highway Remote Scale Characteristic Distribution of cover types throughout study site Landscape Habitat Remote Distribution of slope classes throughout study site Landscape Habitat Remote Distribution of aspect classes throughout study Landscape Habitat Remote site Average size of cover type patch Landscape Habitat Remote Average size of slope class patch Landscape Habitat Remote Average size of aspect class patch Landscape Habitat Remote Average complexity of cover type patch Landscape Habitat Remote Average complexity of slope class patch Landscape Habitat Remote Average complexity of aspect class patch Landscape Habitat Remote Line-of-Site (m) along roadway Local Highway Field Distance (m) to nearest woody vegetation at the Local Habitat Field roadside Line-of-sight to roadway 20 m froin the roadside Local Habitat Field Distance to nearest woody vegetation 20 m from Local Habitat Field the roadside Minimum and maximum slope at roadside Local Habitat Field Distance to nearest center of human activity Local Habitat Field Distance to nearest side road Local Habitat Field Distance to nearest drainage Local Habitat Field Distance to nearest roadside barrier Local Highway Field Distance to. the forest edge Local Habitat Remote Location ofCZ relative to underpass Local Highway Field Cover types associated with CZs compare to Local Habitat Remote available cover types Slope classes associated with CZs compare to Local Habitat Remote available slope classes Aspect classes associated with CZs compare to Local Habitat Remote Available aspect classes "'Measured in the field or from remote photography Landscape Scale Variables Existing research pertaining to wildlife/highway interactions, reviewed in Chapter 2, suggests three main landscape-scale habitat characteristics that influence how wild animals move through the landscape, and consequently, where they approach the roadside. These characteristics include topography (e.g., Romin and Bissonette 42

PAGE 58

1996;.Finder et al., 1999, Alexander and Waters 2000, Hubbard et al., 2000); avoidance ofhuman activity (e.g., Rodriguez et al., 1996a; Clevenger and Waltho 2000); and the continuity and heterogeneity of cover types (Haas 2000; Hubbard et al., 2000). Topography, as defmed by the studies cited above, encompasses the general slope and aspect of a landscape as well as the landforms it contains, such as drainages and ridgelines. Slope and aspect are important because most species avoid excessively steep slopes, and some species may have a preference for a certain aspect. Additionally; topographically complex landscapes (i.e., those with many change in slope and aspect over a small area) may also be difficult to travel through. According to the studies cited above, drainages influence how animals move through the landscape and consequently where they come to the roadside. It is unclear if that is because they act as guideways due to their linear form or animals prefer the habitats available along them. Landscape-scale continuity and heterogeneity of cover type may also influence where animals will approach the roadside, but their role is unclear. Hubbard et al. (2000) found that deer w:ere more likely to approach the roadside in areas with large, contiguous blocks of cover. However, in a simple model of animal movement through habitat patch mosaics, Stamps et al. (1987) showed that the rate of movement should be highest in mosaics comprised of many patches with high edge-to-size ratios, as long as the edge is moderately "soft". Edge softness is defmed by a species' ability to move between the habitats that the edge separate. Hubbard et al. (2000) also suggested that distinct edges between different cover types may act as linear guideways, much like drainages and ridgelines. This intriguing idea qoes not appear to have been addressed in the habitat connectivity 43

PAGE 59

literatUre, nor has it been a topic of edge-effect research (reviewed in Lidicker, 1999). Based on the literature cited above, I chose two types of landscape-scale variables. First, I measured the composition of the landscape, based on the amount of each cover, slope, and aspect class surrounding the highways. Then I measured complexity of the landscape. Metrics of landscape structure that can be used to assess complexity include patch size and patch edge-to-size ratio (ESR). A landscape made up of many small patches is more complex than is a landscape made up of a few large patches. Additionally, a group of patches with small average sizes and long edges relative to their size represent a more complex landscape structure then a group of patches with large areas and shorter relative edge lengths. I did not quantitatively assess the effect of human activity or linear guideways such as drainages, at the landscape-scale. I purposefully chose study sites surrounded by public land. Therefore, most of the human activity that occurred at these sites was limited to the roadside and I considered avoidance of human activity a local-scale variable. I limit my examination of linear guideways to a qualitative assessment because I had no objective methodology for identifying what constituted a linear feature. Local-Scale Variables Existing research pertaining to interactions, reviewed in Chapter 2, suggests a suite of local-scale habitat characteristics that influence how animals use roadsides. These variables include the presence and location of roadside barriers, food, and preferred habitat (see Table 2.1). Avoidance ofhuman activity 44

PAGE 60

and topography, specifically drainages that intersect the roadway, are also thought to influence how animals use roadsides (see Table 2.1). The form, presence, and absence of some of these roadway and roadside habitat characteristics are under the direct control of highway designers and builders. These variables include the amount of cover left in or planted in the right-of-way (ROW), the distance from the roadside to cover, placement of roadside barriers, and the location and configuration of crossing structures, such as underpasses. Based on the information discussed above, I chose local-scale variables that described the relationship of topography, cover, roadside barriers, underpasses, and human activity to locations that animals used to cross roads. I excluded measurements related to the presence of food and preferred habitat because I wanted to focus on variables that all mediumto large-sized species were likely to perceive, and these variables are species specific. Study Site Descriptions I selected two study sites in Southern Rocky Mountains of Colorado, USA. One was located along US 24 at Trout Creek Pass and the other along I-70 at Vail Pass. My primary study site selection criteria included: reasonable proximity to Denver, CO, my home base; the presence of a suitable roadside tracking medium year round; and a wide shoulder to provide a safe work environment. Another consideration was adjacent public land, to reduce potentially confounding effects of human disturbance associated with homes and businesses and to help ensure a reasonably local wildlife population. 45

PAGE 61

Trout Creek Pass The Trout Creek Pass study site (TCP) was located predominantly in Chaffee County, CO, USA and encompassed 11.0 miles (17.8 km) of narrow, two-lane highway that simultaneously serves as US 24 and US 285. The mile postings (MP) in this stretch of road follow the sequence ofUS 24 and range from MP 216.0 approximately 2 miles east of Johnson Village, to MP 226.0, approximately one mile east of Trout Creek Pass (Figures 3.1, 3.2). A small section of the study area, to the east of the Pass, was located in Park County. US 24 is a two-lane road throughout the study area, except for the east side of the Pass where a climbing lane creates a short section with three lanes. Lanes are 3. 7 m wide and shoulders are unpaved. The average annual daily traffic volume is 4000 vehicles (CDOT 2000). Although US 24 is classified as an east-west highway, in the north end ofTCP it runs predominately north-south (MP 221-226). The terrain in this part of the study area was rolling, and cover type consisted of open grasslands and shrub communities west of US 24, and mixed coniferous forests to the east. The shrubs at TCP were tall, and like the forest cover types, they provided good cover. The terrain in the south end of the study area was rugged and highly dissected by dry washes and rocky outcrops, and the vegetation consisted of coniferous stands intermixed with aspen, deciduous shrubs, and small, open grassy areas. Additionally, a narrow shrubby riparian zone associated with Trout Creek, which paralleled the highway in the southern section ofthe study area. Elevations in the study area range from 2830 m at the Pass to 2420 m at MP 216, and the main source of human disturbance, apart from the highway itself, were about 20 homes located mainly in the southern end of the study area. 46

PAGE 62

Cbaffee County Trout Creek Pass Study Site N i Figure 3.1 The location of the Trout Creek Pass study site (TCP) in Chaffee County, Colorado. 47

PAGE 63

N North end of study area South end of study area Trout Creelc Pass Chaffee County Figure 3.2 Detail of the Trout Creek Pass area, location of TCP. 48

PAGE 64

US 24 intersected six major drainages in the study area, which were bridged by large, three-chambered concrete box culverts with concrete floors or by smaller bridge structures with natural floors. Numerous smaller drainages in the study area were bridged with pipes or culverts ofvarious sizes, ranging from 0.03 to 1.0 min diameter. Because snow cover in the study area was usually temporary, this area acted as both summer and winter range for mule deer and elk, although elk were more common at the north end and in winter. Other common terrestrial species included red fox (Vulpes vulpes), gray fox (Urocyon cinereoargenteus), coyote (Canis latrans), mountain lion, bobcat, long-tailed and short-tailed weasel (Mustela frenata, M erminea), and mountain cottontail (Sylvilagus nuttallii). The tracking medium at TCP consisted mainly of native roadside dirt, sand, and gravel, which provided an inconsistent surface throughout the study area. Some locations had very fme-grained dirt along the shoulder and readily took impressions from both hoofed and pad-footed animals. Most locations however, had a mix of dirt, sand and gravel and recorded hooves far more efficiently. The variability in the tracking medium was exacerbated by variations in its ability to retain moisture. Areas with more gravel dried more quickly and yielded poorer quality track impressions. In most places, the unvegetated area along the roadside was about a meter wide. Vail Pass Straddling Summit and Eagle counties, CO, USA, the Vail Pass study site encompassed 12.0 miles (19.4 km) ofl-70, from MP 183.0 to MP 195.0 (Figures 3.3, 3.4). Because of the heavy winter snows this site receives, the suite of animal species present (described below) and their behavior differs substantially between the times of the year when snow is present compared with when snow is not 49

PAGE 65

Eagle County Vail Pass Area Figure 3.3 Location ofthe Vail Pass (VP) and Vail Pass Snow (VPS) study sites, straddling Eagle and Summit Counties in Colorado. 50

PAGE 66

\' To the Town \ ofVail MP 184 ""--J Eagi:UC:unty \ .. _.., ........ ,_ \ t N Summit County and East side of the study area Figure 3.4 Detail of the Vail Pass area, location ofboth VP and VPS, including the location of Copper Mountain Resort (CMR). 51

PAGE 67

present..Therefore, I considered the Vail Pass area to be two different sites, based on snow depths. I will refer to the site as Vail Pass (VP) when the ground was snow-free and as Vail Pass Snow (VPS) when there was snow on the ground . The highway at VP and VPS is designated as an east-west road and the average annual daily traffic volume is 15,500 vehicles (CDOT 2000). Vail Pass, located at approximately MP 190, divides the study area into an east side and a west side. The section ofi-70 encompassed by the study site contained two 4.1 m-wide lanes and 4. 7 m of associated paved shoulders for a total of width of 12.9 m in each direction. The alignments of the east and westbound lanes were independently sited and varied in location and elevation. The median separating the east and westbound lanes varied in width from less then a meter in some places on the west side of the Pass, up to 260m on the east side of the Pass. To a large extent, the natural cover and topography were maintained within the wide median area on the east side. On the west side, Jersey barrierS separated the east and westbound lanes in locations where they were at the same elevation. Stepped retaining walls were used to separate the lanes in locations where one lane was at a higher elevation then the other. On both the east and west sides, additional Jersey barriers were used along the outer shoulder in locations where steep drop-offs occurred. In general, there were far fewer vertical roadside and median barriers on the east side than on the west side of the study site (Figure 3.5). I70 intersected 18 large drainages in the study area, and bridges spanned 11 of them. These bridges provided high quality highway crossing opportunities for wildlife as the drainages they spanned are wide (up to 230m), and the natural cover below most was largely undisturbed. Most of the smaller drainages that intersected the roadway in the Vail Pass area were diverted into pipes less then 1 m in 52

PAGE 68

East Side of the Pass --1-70 --Barriers --Underpasses Figure 3.5 The locations of all barriers and underpasses on 1-70 at Vail Pass. Note the distribution ofbarriers and the alignment ofthe underpasses on the two sides of the Pass. 53

PAGE 69

diameter. The primary cover type in the V:ail Pass area was mixed coniferous forest interspersed with aspen stands, sub-alpine meadows, and willow carrs. These shrubby willow areas did not provide good cover. The elevation of the study site ranged from 2730 to 3165 m, and sources of human-induced disturbance, aside from the highway itself, included a rest area, truck tum out and maintenance shed at the summit and the Copper Mountain Resort at the base of the east side. Common terrestrial wildlife species in this study area included red fox, bobcat, mule deer, elk, and mountain lion during the snow-free months. Snowshoe hare (Lepus americanus), coyote, long-tailed and short-tailed weasels, and American marten (Martes americana) were present year-round. The tracking medium at VP consisted mainly of road sand left over from the previous winter. It provided a uniform tracking surface throughout the study area. The sand was a mix of fme and coarse grains, and took the imprint of hooves readily, but was less efficient at recording the passage of pad-footed animals. Frequent thunderstorms kept the roadside at VP relatively moist throughout the summer, further improving its quality as a tracking medium. The thick layer of road sand also choked out most roadside plants, and an unvegetated swath of sand generally extended from 1-3m away from the paved shoulder. At VPS, I used snow as the tracking medium. Data Collection Methods: Tracking Standard Tracking Methods I recorded locations throughout all three study areas where mediumand large sized mammals (mule deer, elk, coyote, fox, bobcat, mountain lion) crossed the as indicated by their At TCP and VP I checked roadside transects 54

PAGE 70

200 IIi. in length -for tracks during each field session. To ensure that transects were distributed throughout a study area and did not overlap, I used a stratified random selection approach, varying transect location for each data collection session. At each transect, a field assistant or I walked along the highway at the pavement's edge and looked for tracks left in the unpaved shoulder. At TCP, traffic was light and I crossed the highway to walk along both sides of it, and recorded tracks from both sides. AtVP however, I only walked along the outer edges of the westand east-bound lanes. Due to high traffic volumes and speeds, I considered crossing the highway to access the median-side roadside unsafe. I recorded track locations using a handheld GPS device\data logger (Geo Explorer II, Trimble) that automatically recorded location while I entered information through a menu-driven interface. All tracks of the same species observed within a 5-meter stretch were recorded as a single track record (TR). Each TR contained the following information: species of animal, number of animals, location (UTM coordinates), activity (described below), surface (dirt or snow) and date. I downloaded data files from the data logger and used Trimble's proprietary software to convert them to Excel spreadsheet and Arc View shapefile formats for analysis. I interpreted the activity of an animal from the pattern of tracks it left behind. Activity included four classifications: I) Crossing -a track pattern indicating the animal passed across the roadway from one side to the other; 2) Approach -a track pattern indicating the animal approached the roadway but did not pass across it to the other side; 3) Parallel -a track pattern indicating the animal walked along the for a distance of 10 m or more; and 4) Undefmed -a track pattern that did not clearly indicate any of the three behaviors described above. At TCP, I only classified a TR as "crossing" if a matched set of tracks was found on both sides of the road. At VP, I did not confirm crossings in this way because I considered it 55

PAGE 71

unsafe to cross the roadway. Instead, I classified track sets that were perpendicular to the road and did not have a matched set within 20 m going in the opposite direction as "crossing." Because the swath of tracking medium along the roadside was wide at VP, it was usually possible to "read" an animal's behavior at the roadside quite clearly and I only designated a TR as "crossing" when I was reasonably sure the animal had indeed passed to the other side of the roadway. At TCP, I collected data twice weekly from September through June, weather permitting. I collected track data only once a week during July and August because animals moved out of the area during summer, and very few tracks were found during this period. Data were collected along 11 randomly chosen 200 m transects and existing roadside substrates (fine-grained dirt, sand, or snow) used as a tracking medium. At VP, I collected data using the protocols described above from June through November. Data were collected twice weekly, weather permitting, along 10 randomly chosen 200m transects and existing roadside substrates (fme-grained dirt or sand) used as a tracking medium. Standard Underpass Monitoring In addition to monitoring the roadside for tracks, I monitored some highway structures (bridges, oversized concrete box-culverts) at both study sites that could have been used by animals to cross under the highway. Although only one of the monitored structures was constructed specifically to act as a highway underpass for wildlife, I will refer to all these structures as "underpasses." All underpasses monitored spanned either narrow perennial streams or intermittent drainages that only carried water during spring run-off or during storm events, and offered plenty of dey substrate for animals to use when they passed through. I created track beds from locally available sand and soil at both ends of each monitored structure. An 56

PAGE 72

animal was recorded as passing through a structure only when I observed a matched set of tracks at both ends. At TCP, a field assistant or I monitored a subset of 10 underpasses, chosen based on accessibility and safety considerations. Due to time constraints only two underpasses were randomly chosen and checked during each data collection session. Track beds and the roadside within 100 m of either side of the structure were checked for tracks to determine if animals crossing at that location had crossed at-grade or used the structure. Additionally, I checked the track beds, but not the adjacent roadside, in the structures located in the drainages at :MP 215.0, MP 216.5, Shields Gulch (MP 218.4), and Magee Gulch (MP219.2) as often as time permitted. I checked these four large drainages often because I assumed they were most likely to be used, and I wanted to record the full variety of animals willing to cross under the unfenced road at TCP. At VP, I monitored four of the 17 underpasses; the other 13 were deemed unsuitable for monitoring due to safety considerations, high levels of human use, or excessively large size, which made maintaining the trackbed difficult. I checked track beds and the roadside within 50 m of either side of each underpass for tracks to record if animals crossing at that location had crossed at-grade or used the structure. All four underpasses were checked for tracks as a part of every data collection session. Snow Tracking Methods Snow is a superior tracking medium. Snow cover allows tracks to be spotted much more easily and allows the tracks of a wider range of species to be observed. In addition to the tracks of large species that were observed when using roadside dirt/sand as tracking medium, I recorded smaller species such as snowshoe hare, cottontail rabbit, fox, weasel, and marten from snow. Snow tracking was conducted 57

PAGE 73

at VPS. December through March during both 2000/2001 and 200112001. I did not implement snow-tracking protocols at TCP because sufficient snow cover at this site was infrequent, unpredictable, and ephemeral. Thus, even on the few occasions when there was snow on the ground at TCP, the standard tracking procedures described above were followed. However, the snow did present the opportunity to record fox and rabbit tracks at TCP, which rarely left readable tracks at the roadside otherwise. Using snow-tracking methods, I observed the entire VPS study area, as opposed to a subset of transects, for tracks. Due to the snow depths, far fewer animals are present in the Vail Pass area during winter than during the summer. Thus, finding and recording all trails present was a reasonable task. A field assistant or I located all animal trails that entered the roadway within the study area by driving slowly ( < 25 km/h) along the shoulder. When a trail was observed it was identified by species, and crossing success determined. Using the GPS device/data logger, I recorded all information, including species, activity, number of individuals, substrate, and the UTM coordinates of the trail's intersection with the highway. Additionally, I monitored five underpasses for animal use during the winter, as weather permitted. The trails of all animals entering the space underneath these structures were observed in the and each animal's behavior was recorded as either passing through or not passing through. The number and timing of snow tracking sessions was weather-dependent. I only conducted snow tracking after a recent snowfall. Storms occur roughly every four to eight days over Vail Pass during the winter. High winds, warm temperatures, and sunny days, all of which can render tracks unreadable, are common ()n Vail Pass throughout the winter, so it was necessary to track as soon as possible after a storm. I conducted most roadside tracking sessions 6-32 hours after a snowfall. The terrain 58

PAGE 74

under an underpass and the underpass itself protects tracks from wind and sun. Therefore, I waited to check underpasses for tracks until after roadside tracking was completed, usually 32-58 hours post snowfall. Depending on the number of tracks located, one to two days were required to conduct the roadside tracking, and structure monitoring required an additional day of work. I did not conduct snow tracking and structure monitoring while snow was falling. Occasionally there was insufficient time between storms for a complete 2to 3-day data collection cycle. Additionally, due to heavy traffic volumes, roadside work was deemed unsafe on weekends and holidays, and only structure monitoring was conducted on these days. Animal Abund3nce To determine how local animal densities might influence crossing rates, I attempted to estimate the relative abundance of animals in the immediate vicinity of the roadway at both study sites. Within 36 to 60 homs of a snowstorm, I walked transects 1.6 km long parallel to and 300 m from the roadside, and counted all animal trails that intersected my path. Because other snow tracking tasks also had to be preformed in this timeframe, it was difficult to schedule these off-road transects. Additionally, these transects usually required breaking trail in deep snow over rough terrain. Therefore, only one or two could be walked in a single day. Because of these constraints, only five off-road transects were walked at VPS and 14 atTCP. In addition to collecting field data, I also contacted the Colorado Division of Wildlife (CDOW) and obtained statewide mule deer and elk population estimates. The CDOW's data reports population estimates by Game Management Units (GMUs), areas that the DOW believes are relatively homogenous in both the 59

PAGE 75

habitaf quality and the population density of the species of interest. The boundaries of these GMUs differ for mule deer and elk. Although the GMUs are many times larger than my study sites, these population estimates provided a useful indication of the relative densities of the mule deer and elk between the two sites. Data Collection Methods: Habitat Measurements Landscape-Scale Habitat Measurements I made all measurements of landscape structure for landscape-scale comparisons from digital data layers, using the Arc View software package. These data layers were generated from aerial and satellite photography, and I used coverages created by the U.S. Geological Survey (USGS) when available, including the National Land Cover Data (NLCD) for TCP, and 10-meter contour resolution Digital Elevation Models (DEMs) at VP and VPS. Because my field experience suggested that the NLCD coverage of the Vail Pass area misclassified the cover type of large areas, I used a digital vegetation map created by the Forest Service for vegetation measurements at VP and VPS. For the TCP area, 10m resolution DEMs were not available from the USGS. I commissioned the Remote Sensing and Geographic Information Group (RSGIG) located in the Denver, CO, office of the U.S. Bureau of Reclamation to create a 10m OEM of the TCP site specifically for my project. This OEM was created by combining survey control points of known elevations with existing stereo photography using the ERDAS Orthobase software package. At all three study sites, the landscape I measured was the area encompassed by the ridgelines that provided visual boundaries surrounding the highway. I derived the vegetation patterns of these landscapes from the digital cover maps and the topographic patterns from the DEMs. The NLCD is divided into 21 cover type 60

PAGE 76

classes in Colorado, and I reclassified the Forest Service data layer to match those classes. I used the Arc View extension Spatial Analyst to process the DEMs into 19 slope classes and nine aspect classes. I classified topographically level areas as "flat," then divided slope by 5-degree increments for 18 additional classes and aspect by increments of 45 degrees for eight additional classes (Table 3.2). Table 3.2 Definitions used to divide cover. slope, and aspect classes on digital maps of the three study sites Digital Coverage Class Definitions Type Cover Type Water Slope Aspect CommerciaVfransportation Barren (rock or sand) Quarries Deciduous Forest Coniferous Forest Mixed Forest Shrub land Grassland/Herbaceous Emergent Wetlands Woody Wetlands Flat 0-5 46-50 6-10 51-55 11-15 56-60 16-20 61-65 21-25 66-70 26-30 11-75 31-35 76-80 36-40 81-85 41-45 86-90 Flat North Northeast East Ice and Snow Low Density Residential High Density Residential Transitional Orchard Hay Row Crops Small Grains Fallow Urban Recreational Southeast South Southwest West Northwest I was careful to choose digital base maps that adequately reflected actual land cover classes and their boundaries. I based my assessment on my familiarity with the study sites. I wanted to be confident that the maps reflected variations animals 61

PAGE 77

could perceive and respond to. The NLCD is divided into 21 classes in the Colorado, 10 of which are naturally occurring land cover types (Table 3.2). Because these classes are broadly defmed, I believe mediumto large.,.sized mammals can readily perceive the variation they represent, and this classification scheme was reasonable for my analysis. Similarly, I chose the slope and aspect increments (Table 3.2) because my experience with animal behavior and habitat selection suggests that animals respond to topographic variation at those levels. Local-Scale Habitat Measurements I collected two types of local-scale habitat data in the field from both study site locations. The first type of local-scale data that I collected was the characteristics of point locations within crossing zone (CZs) as well as the characteristics of random points for comparison. The methods I used to identify CZs are described in the Data Analysis section of this chapter. I point locations within the CZs by visually inspecting maps of the CZs, and locating points where the highest density of crossing TRs occurred. I chose random point locations with Arc View, using a script that generated random points along lines representing the highways in the study areas. Using Arc View, I determined the UTM coordinates of both random and crossing points and identified the corresponding location in the field using the GPS device. I then used the data logging capability of the GPS device to record habitat measurements for each point. Finally, I downloaded these data in spreadsheet format to conduct statistical comparisons between the CZ points and the random points. At each point measured, I recorded information from both the pavement's edge and 20 m from the pavement edge, to reflect what an animal would see as it approached the road as well as when it began crossing. At TCP I took measurements from both 62

PAGE 78

sides of the road, resulting in a pair of measurements at each point. At VP, I treated the westbound and eastbound alignments separately, locating points to measure independently along each alignment. I recorded data from only the outer edge of the alignment as I deemed it unsafe to cross over to the median-side edge. The variables that I measured and recorded at each point are described in Table 3.3. The line-of-sight (LOSAR, LOSTRa, LOSTRb) and distance-to-cover measurements (DWV, DWV20) were made from the perspective of a medium-sized animal (e.g., coyote, bobcat, mountain lion) as well as from the perspective of a large-sized animal (deer or elk). These two "animal's eye views" of the habitat were simulated by placing a laser rangefmder on a monopod sized to elevate the eyepiece 0.8 or 1.6 m above the ground. Line-of-sight distances and distance-to-cover, as observed through the rangefmder, were then recorded at both monopod heights. The second type of data I collected in the field was the locations of features along the roadside, including bridges (representing the locations of_both underpasses and drainages) and roadside barriers (cliffs, walls, guardrails, Jersey barriers). I used these data to create digital data layers for use with Arc View. I collected these data using the GPS device's setting for recording continuous data along a line. Using a roof-mounted antenna, I drove slowly (20-25 kmlh) along each feature of interest and collected positions for the entire length of each feature, then converted the positions into Arc View shapefiles using Trimble's proprietary software package. I used these data layers to measure the distance of CZs to their respective features. I 63

PAGE 79

Table 3.3 Variables measured in the field at crossing zones (CZs) points and random points for comparison of local-scale habitat charateristics At Roadside 20m From Roadside Variable Measured Abbreviation Line-of-Sight (m) along roadway at LOSAR 1.6 m, 0.8 m Distance (m) to nearest woody DWV vegetation at 1.6 m, 0.8m* Distance to nearest side road DTSR Distance to nearest center of human DTHA activity Line-of-sight to roadway at 1.6 m and0.8m* LOSTRa*** LOSTRb*** Notes Two LOSAR measurements were taken at each point, one in either direction. Measurements were taken along eight equidistant radii originating from the center of the point being measured, at both heights. Distance to nearest visible side road Distance nearest house, rest area, business, or maintenance depot. View to road from point being measured, categorized as 0 = completely obstructed, 1 = partially obstructed, 2 = unobstructed. Distance to nearest woody vegetation at 1.6 m, 0.8m* DWV20 Eight measurements were taken along eight originating radii emanating from the center of the point, at both heights. Minimum and maximum slope between roadside and 20 m from road side MINSL, MAXSL Measured in degrees These measurements were taken at both 0.8 m above the pavement surface and 1.6 m above the pavement surface No values were recorded for radii that intersected rock or dirt. Woody vegetation 70m or farther from the point was all recorded as 70 m. Titree observations were taken at each point, one looking straight to the road (LOSTRa), and then two more, looking to the road 45 either side from the line of the first observation. The two side views were averaged (LOSTRb ). In addition to making field measurements of the roadside habitat, I also used aerial photos and existing digital data layers to make some local-scale measurements. I digitized lines representing the forest boundaries from aerial photos, then used that data layer to measure the distance of CZs to the nearest forest edge with Arc View. Additionally, I used the digital vegetation and topographic data layers described in the Landscape-Scale Habitat Measurements section to compare the cover, slope and aspect classes associated with CZs to what was available along the entire roadside. 64

PAGE 80

Data Analysis: Identifying Patterns Descriptive Summazy of Track Records I considered the data in three separate sets, based on study site and snow depth: Trout Creek Pass (TCP), Vail Pass snow-free (VP), and Vail Pass with snow cover (VPS). I calculated the total number ofTRs recorded at each site, counted them by activity and species, and created maps depicting the TR locations. I further divided these primary data sets into subsets, as described below, based on the spatial patterns of the TRs within each study site, and landscape characteristics of the study sites. This approach was necessary because visual analysis of the maps indicated that the density of TRs throughout each study site was uneven. Identifying First Order Patterns The maps suggested that both frrst order (large-scale) and second order (small scale) patterns were present. Because variations in frrst order patterns can mask or swamp second order patterns, small-scale spatial patterns must be studied over scales at which the first order effects remain homogenous (Baily and Gatrell 1995, p. 9l):.Therefore, I divided each of the three primary data sets into sub-areas, based on the extent of homogenous first order patterns. I used visual analysis and simple counts of the crossing TRs at each sttidy site to demark their frrst order patterns. I confrrmed the presence of these first order patterns by using SPSS v.IO.l for Windows to perform chi-square tests to determine if the observed distributions between sub-areas of apparent high and low density TRs did indeed differ from an even distribution. 65

PAGE 81

Identifying Second Order Patterns (Crossing Zones) I sought to confirm the presence of second order (small-scale} patterns within the sub-areas by looking for groups of TRs that were closer to one another than expected. For this analysis I used only those TRs that had been classified as "crossing" in the field so that I could use designated clusters as indicators of CZs. I defined a TR as part of a cluster if the median distance to its n nearest neighbors was less then expected by chance. The median, rather than the mean, distance between a point and its n nearest neighbors was chosen as the metric of comparison for identifying CZs because the distribution of nearest-neighbor distances was skewed to the right. The median is less influenced by outliers, providing a more conservative estimate of the data's central tendency. Because frrst order patterns affect the expression of second order patterns, the number of nearest neighbors I considered varied with the total number of TRs in each sub-area. For sub-areas with fewer then I 00 TR, I considered the median distances of a point to its three nearest neighbors. For sub-areas with 10 I to I99 TR, I considered the median distances of a point to its five nearest neighbors, and for sub-areas with 200 to 299 TR, I considered the median distances of a point to its seven nearest neighbors. I defined an n nearest-neighbors distance smaller than three standard deviations from the expected median distance as not occurring by chance, and therefore, an indication of clustering. To determine the expected nearest neighbors distance for each data set, I used a Monte Carlo (Besag and Diggle, I977) approach. I distributed points randomly along a line representing the roadway of interest to simulate a possible distribution of crossing TRs along it, and then measured the n-nearest-neighbors distance for each point. I used 1 000 simulations to generate the expected distribution of nearest-66

PAGE 82

neighbors distances. A script written for Arc View in the A venue programming language (Martin, 2001) was used to perform all simulations automatically, based on user-defmed input parameters. Using this script, I specified a line, the number of points to use, the number of nearest neighbors to measure, and the number of simulations to run. The script then automatically performed the number of simulations requested, calculated the mean, median, maximum, and minimum n nearest-neighbors distance of each simulation, and stored these values in a spread sheet. After the simulations were completed, I exported the spreadsheet to Excel and calculated summary statistics of each value. Measuring Crossing Zones After identifYing the TRs with n closer-than-expected nearest neighbors on a map of all crossing TRs, I buffered them with a radius the length of the expected median n-nearest-neighbors distance, and dissolved the boundaries of all overlapping buffers. I chose this buffer size as it should include all the points that contributed to a given TR's nearest neighbor measurement, on average. I designated all crossing TRs contained within each buffer as part of that CZ (Figure 3. 7). I then measured the length of each CZ as the distance between the two outer-most TRs in Arc View. They are accurate to 10 m. A group of TRs was only considered a CZ if it contained at least as many TRs as were used in that data set's nearest neighbors calculation (e.g., three nearest neighbors required considering four TRs). Finally, I inspected the TRs that comprised a cz to determine if they represented independent events. TRs of the same species recorded the same day and within 50 m of each other were not considered to be independent. 67

PAGE 83

Underpass Use I calculated the frequency of underpass use in two different ways. First, I compared the number of times at least one set of tracks, of any species, was recorded as passing through to the total number of times the structure was checked for tracks. Although it was common for me to fmd multiple track sets each time I checked an underpass, I could not determine if one animal passing through an underpass many times or many animals passing through once created the multi-track track sets. Because a group of n animals crossing together may be regarded as one crossing event, rather then n independent events, I grouped these data. However, because the firSt approach lumps data, important information about species use is masked. Therefore, I also compared the number of times at least one individual, by species, was recorded as passing through to the number of times the structure was checked for tracks. For example, ifi checked an underpass.and recorded a bobcat and three rabbits as passing through, I counted the underpass as used once by bobcat and once by rabbit. Using this counting method, it is possible that I could record more through passes then the number of times the underpass was checked. Therefore, I report the frequency of use calculated with this approach as a ratio. Finally, at VPS I also counted the total number of animals passing through the underpasses. Snow is an excellent tracking medium, aild few animals are present in .the Vail Pass area when the snow is deep, making it possible to accurately count every trail. 68

PAGE 84

Data Analysis: Quantifying Relationships Quantifying the Relationship of First Order Patterns to Landscape Structure As detailed previously, I split each of the three study sites into two sub-areas, based on the frrst order patterns of TR density along the roadside. I evaluated the influence of the surrounding landscape on this first order pattern of TR distribution by comparing the landscape structure of the each study site's two sub-areas. Metrics of landscape structure that I considered included the average size and ESR of patches defmed by different classes of cover type, slope, or aspect in each study site sub-area. I used the Arc View extension Patch Analyst to count the total number of patches created by variations in cover, slope, or aspect, as well as to calculate the area of each patch and the length of its perimeter. I defmed a "patch" as a contiguous area comprised of a consistent class. Although some of the samples I compared were neither normal nor homogeneous in their distributions, I chose a tWo-sample t-test (SPSS 10.1) to compare the mean values of each variable. SPSS offers a correction for heterogeneous distributions, and my samples size were very large, relaxing the need for normal distributions. Quantifying the Relationship of Second Order Patterns to Local-Scale Features The approach I used to quantify the local-scale habitat characteristics associated with crossing zones varied according to type of feature under consideration and the data's source. For variables measured in the field (listed in Table 3.3), I calculated the average of each variable measured within CZs and the average value meaSured at random points. I then compared these two values to determine if there was_a differe_nce between CZs and random locations. Because these data were non69

PAGE 85

normal and often had significantly different variances, I used the non-parametric Mann-Whitney U test (SPSS 10.1.0 for Windows) to determine if the means differed from one another. For the cover, slope, and aspect variables measured on GIS data layers, I used the Arc View extension Patch Analyst to calculate the total area of each cover, slope, and aspect class within 100 m of the CZs and within 100 m of the entire roadway. I used Excel spreadsheet functions to run a chi-square test to compare the proportion of cover, slope, and aspect classes contained in each of the paired data sets. I also used a GIS-based approach to evaluate if the distribution of CZs in relationship to drainages, roadside barriers, and the forest boundary deviated from what would be expected. I used scripts written for Arc View (Martin 2001) in the A venue programming language to measure the actual distance of each TR to the feature of interest, and to implement a Monte Carlo simulation (n = 1 000) to generate an expected distance for comparison. A single simulation consisted of distributing points randomly along a line representing the road, measuring the distance of all points to the nearest feature of interest, then calculating the average point-to-feature distance. The number of points used corresponded to the number of TRs that made up the CZs in the study site sub-area l;>eing evaluated. The script carried out the Monte Carlo simulations automatically, based on user defmed input parameters, including Arc View shapefiles representing the feature of interest and the roadway, the number of points to use, and the number of simulations to run. The script then executed the number of simulations requested, calculated the mean, median, maximum, and minimum distance, and stored these values in a spreadsheet. After the simulations were complete I exported the spreadsheet to Excel and calculated the summary stati:;tics of each value. I used a 70

PAGE 86

two-sample t-test, implemented with Excel spreadsheet functions, to compare the actual mean TR -tofeature distance to the expected mean. If the two means were not significantly different at a = 0.05, I concluded that the TRs were randomly distributed in relationship to the features of interest. Additionally, the script counted the number of points in each simulation that were placed adjacent to a barrier so I was able to calculate an expected value. I compared the proportion of TRs that were expected to be adjacent to a barrier to the actual proportion ofTRs located adjacent to a barrier with a chi-square test. To examine if the presence of underpasses had any effect on the locations where animals crossed at-grade, I graphed the relationship between meters of underpass along a kilometer of highway against the number of crossing TRs along that kilometer ofhighway. I estimated both values using a moving windows analysis. Each window along the roadway was one km in length and was shifted in 100 m increments. Quantifying Animal Abundance The small data sets from the off-road transect collected at both TCP and VPS do not provide a statistically valid sample to evaluate and compare animal densities. Likewise, the gross scale of the CDOW data does not provide an accurate estimation of the animal populations at the roadside specifically. Therefore, I summarized these data by simply counting the number of tracks observed and calculating the average number of tracks/transect at each study site, and reporting the CDOW deer and elk population data. 71

PAGE 87

Summary To determine if the locations where animals cross the highway are different from random locations, I chose variables for comparison by reviewing the current literature and considering which roadway and roadside features highway designers' control. To collect data, I chose two study areas in the Southern Rocky Mountains. First, I recorded where wild-animals crossed the road. Next, I measured characteristics of the habitat both directly in the field and from digital data layers that created from field measurements or remote photography. Depending on the type of data collected, I analyzed the data by comparing average values, using a Monte Carlo approach to generate an expected distribution to compare to actual distributions, or by comparing used to available habitat. I also collected data about underpass use and summarized it with simple counts. 72

PAGE 88

CHAPTER4 WILDLIFE HIGHWAY-CROSSING ZONE INVESTIGATION: FINDINGS Introduction In the following chapter I report the results of my data collection and analysis efforts. The discussion is organized according to the order followed in Chapter 3. Simple summaries of the track record (TR) data are reported first. Results of the analyses that correlated firSt order TR patterns to characteristics of the surrounding landscape are followed by results of the analyses that correlated second order TR patterns to characteristics of the roadside. Finally, I summarize the underpass use and the animal ablUldance trends at all three study areas . Descriptive Summary of Tracks I collected track data 130 times at Trout Creek Pass (TCP) between 28 January 2000 and 4 July 2001, recording a total of535 TRs, representing 832 individual animals. I conducted a total of 91 tracking sessions when there was no snow cover at Vail Pass (VP), comprised of 40 sessions during 2000 and 51 sessions during 2001. I recorded a total of778 TRs, representing 1155 individuals, at VP. When there was snow on the ground at Vail Pass (VPS) I collected track data on 18 occasions during 2000/01 and on 30 occasions during 2001/02, for a total of 48 snow tracking sessions, and I recorded a total of771 TRs, representing 978 individuals. The TRs are summarized by species and activity for each study site in tables 4.1, 4.2, and 4.3. 73

PAGE 89

Table4.1 Summary ofTRs by species and travel at TCP Species Crossing Not Crossing (%crossing, SJ:!ecies) by SJ:!ecies) Mule Deer 219(53.0%) 194(47.0%) Elk 40 (71.4 %) 16(29.6 %) Coyote 10 (27.8 %) 26 (71.2 %) Rabbits/Hares 7(70.0%) 3 (30.0%) Fox I (25.0 %) 3 (75.0%) Mountain Lion 1 (100%) Bobcat 1 (100.0 %) Other 2(40.0%) 3 (60.0%) Unknown 7 (100.0 %) Total 278 257 Table 4.2 Summary ofTRs by species and travel at VP Species Crossing Not Crossing (% crossin11 sEecies) (%not crossing, by sEecies) Mule Deer 191 (34.9 %) 284 (65.1 %) Elk 117 (43.2%) 154 (56.8 %) Coyote 8 (28.6%) 20 (71.4 %) Mountain Lion 1 (100.0 %) Moose 1 (100.0 %) Other 1 (100.0 %) Unknown I (100.0 %) Total 319 459 Table 4.3 Summary ofTRs by species and travel at VPS Species Coyote American marten Weasel species Snowshoe hare Redfox Elk Total Crossing (%crossing, by species) 433 (74.5%) 12 (35.3 %) 23 (45.1 %) 39(76.5 %) 507 Not Crossing (%not crossing, by species) 148 (25.5 %) 22 (64.7 %) 28 (54.9 %) 51 (23.5 %) 2(100.0 %) 8(100.0%) 264 Total (%of all TRs, SJ:!ecies) 413 (77.2 %) 56 (10.5 %) 36 (6.7%) 10 (1.9%) 4(0.7 %) 1 (0.2 %) 1(0.2 %) s (0.9 %) 7 (1.3 %) 535 Total {% of all TRs, seecies_L 475 (61.0 %) 271 (34.8%) 28 (3.6 %) 1(>0.1%) 1 (>0.1 %) 1(>0.1%) I (>0.1 %) 778 Total (% of all TRs, by specie 581 (75.3 %) 34 (4.4 %) 51 (6.6%) 90 (11.7 %) 2(0.3 %) 8 (1.0 %) 771 The species recorded most often at TCP and VP was mule deer (77.2 and 61.0 %, respectively), and I also recorded a substantial number of elk at VP (34.8 %). Coyotes were the species recorded most commonly at VPS (75.3 %). The proportion of tracks indicating crossing varied among the three study sites, as did the animal species that crossed most frequently. Animals were most likely to cross 74

PAGE 90

at VPS (65.7% of all tracks) and least likely to cross at VP (41.0 %). The crossing rate at TCP was intermediate (52.0 %). The crossing rate at TCP was not significantly different from either VP 0.10) or VPS 0.05), but animals were significantly more likely to cross the highway at VPS, as compared with VP (l = 9.286, p < 0.05). Distribution of Tracks Animals are more likely to cross highways at certain locations at both the landscape and the local scale. Quantitative as well as visual analyses of the patterns created by the distribution of track records (TRs) along the roadside indicated that both first order and second order clustering existed at all three study sites. I interpreted the second order TR clusters as indicators of locations where animals preferred to cross the road, and designated such locations as crossing zones (CZs). The results of my spatial pattern analysis of the TRs are described in detail below. First Order Patterns Mapping the locations of crossing TRs within their respective study sites revealed that crossing TRs were not evenly distributed throughout any of the study sites. At each of the three sites, I observed more crossing TRs located in a defmable sub area of the site. At TCP I recorded far fewer TRs along the portion ofUS 24 located north of where Trout Creek intersects the highway. The resulting lowand high-density TR areas corresponded with differences in both topography and land cover north and south of Trout Creek, and I divided the data into two sub-areas, north (MP 221.5-226.0) and south (MP 216.0-221.5), accordingly (Figure 4.1). At VP, I recorded less than half as many TRs on the west side of the Pass than the east side. As discussed in Chapter 3, the design of the highway differs substantially between the two sides of the Pass. Therefore, the Pass was a reasonable dividing 75

PAGE 91

line between the differing first order patterns, and I subdivided the data into two sub-areas, east (MP 190.0-195.0) and west (MP 183.0-190.0; Figure 4.2). When snow is present in the Vail Pass area, large, steep snowdrifts along the roadside are created throughout the study site by snowplows. These drifts masked some of the structural differences of the roadway between the east and west sides of the Pass, and it became less of a natural dividing line. Additionally, the distribution ofTRs when snow was on the ground was consistent throughout the VP study area, except for the 2.5 miles of roadway closest to Copper Mountain Resort ski area (CMR). During the months with snow cover I recorded a clearly disproportionate number (75.0 percent of total) ofTRs in the portion ofthe study area closest to CMR. Therefore, I divided VPS into two sub-areas, CMR (MP 192.5-195.0) and Not CMR (MP 183.0-192.5), based on the location of the resort (Figure 4.3). At all three sites the first order patterns, which were apparent when I mapped the locations of all TRs, were mimicked when I mapped crossing TRs only. Additional quantitative analysis supported my initial impressions. A simple count revealed that at TCP nearly 80 % of crossing TRs were located in the southern half of the study area, at VP 68.5 %of crossings TRs were located on the east side of Vail Pass, and at VPS 68.4 % of crossing TRs were adjacent to CMR. Chi-square tests indicated that the proportion of crossing TRs located in the sub-areas departed significantly from either an even or a random distribution, based on the linear distance of the roadway at each study site sub-area (Table 4.4). 76

PAGE 92

N i TCP South Figure 4.1 Locations of all crossing TRs recorded at TCP. Note the higher density of TRs in the south end of the study area. 77

PAGE 93

N i 1-70 Westbound Figure 4.2 Locations of all crossing TRs recorded VP. Note the higher densities of tracks on the east side of the pass. 78

PAGE 94

N i VPSCMR VPSNotCMR Figure 4.3 The locations of all crossing TRs recorded VPS. Note the higher density of tracks near CMR. 79 .

PAGE 95

Table 4.4 Distribution of crossing TRs within the sub-areas of each study site Study Site Sub-Area Number of Linear Length TRslkm Chi-Square Crossing TRs of Sub-Area TCP South 219 8.9km 24.6 l= 19.13, p < 0.00 North 59 7.3 km 8.1 VP East 227 8.8km 25.8 x2 = 80.20, p < 0.00 West 92 11.3 km 8.1 VPS CMR 401 4.0km 100.2 x2= 214.06, p < o.oo NotCf\.1R 106 15.4 km 6.7 Second Order Patterns Visual analysis of the mapped crossing TRs in the sub-areas of each study site suggested that additional small-scale, or second order, clustering ofTRs was present within the first order clustering discussed above (Figures 4.1, 4.2, and 4.3). Nearest neighbors analyses confliiiled my impression of second order patterns at all three study sites. At TCP south the nearest neighbors' analysis indicated that 60.4 % of TRs were more clustered than would be expected by chance. At TCP north 58.6 % ofTRs were more clustered than expected by chance. At VP and VPS, I analyzed clustering separately in the west-:and eastbound lanes. At VP, TRs recorded along the westbound lanes were more likely to be clustered together than those recorded along the eastbound lanes. This was true at both VP east (71.8 versus 47. 8 %) and VP west (66.7 versus 34.4 %) of the Pass. At VPS, 42.9 of the TRs in the westbound lanes and 57.1% ofTRs in the eastbound lanes were closer to one another than expected adjacent to CMR In the rest of the study area, 63.9% ofTRs along the westbound lanes and 72.9% ofTR along the eastbound lanes were more clustered than expected by.chance. 80

PAGE 96

Crossing Zones I identified CZs based on the second order patterns of crossing TRs. I interpreted groups of crossing TRs that were more clustered than expected by chance as an indication that animals focused crossing activity along that stretch of highway. At TCP north I identified five distinct CZs (Figure 4.4). A sixth CZ was discarded because the TRs that defmed it were all the result of crossing by a single herd of elk and could not be considered independent events. The identified CZs ranged from about 80 m to 300 m in length. Distance between them ranged from 240 m to 2630 m. At TCP south, I identified I 0 CZs (Figure 4.5). They ranged from about 30 m to 600 m in length, and the distance between them ranged from 200m to 1120 m. At VP east, I identified five distinct CZs along the westbound lanes and four along the eastbound lanes (Figure 4.6). These CZs ranged in length from 100 to 760 m and the intervals between them ranged from 140 to 151 0 m in length. I dropped three additional CZs along the westbound lanes from consideration because they were defmed by four or fewer TRs, and in each case at least three of the TRs were from the same species, recorded during the same data collection session, and could not be considered independent events. CZs along the westand eastbound lanes were not strongly aligned with one another. At VP west, I identified five distinct CZs along the westbound lanes and three along the eastbound lanes (Figure 4.7). Crossing zones on this side of the pass ranged in length from 50 to 490 m in length and the intervals between them ranged from 650 to 2480 min length. I dropped one CZ along the westbound lanes because all but one of its identifying TRs were recorded during the same data collection session and were from the same species. All three of the CZs identified along the eastbound lanes are strongly aligned with CZs along the westbound lanes. 81

PAGE 97

Figure 4.4 Locations of the crossing TRs that make up the CZs in the north end of TCP. 82

PAGE 98

' Figure 4.5 The locations of crossing TRs that make up the CZs in the south end of TCP 83

PAGE 99

N i Figure 4.6 Locations of crossing TRs that make up the CZs on the east side ofVP. 84

PAGE 100

I-70 Eastbound N I :Eagle County Summit County Figure 4.7 Location of crossing TRs that make up CZs on the west side of VP. 85

PAGE 101

At VPS CMR, I identified seven distinct CZs along the westbound lanes and four along the eastbound lanes (Figure 4.8). Crossing zones in this sub-area ranged in length from 40 to 310 m in length and the intervals between CZs ranged from 70 to 760 min length. I dropped five CZs along the westbound lanes from consideration because they were defmed by fewer than eight TRs. I dropped three additional CZs along the eastbound lanes from consideration because they were defmed by fewer than six TRs. CZs along the westand eastbound lanes were somewhat aligned with one another. At VPS Not CMR, I identified five distinct CZs along the westbound lanes and nine along the eastbound lanes (Figure 4.9). Crossing zones in this portion of the study area ranged from 60 to 530 min length and the intervals between them ranged from 310 to 4290 m in length. I dropped one CZ along the westbound lanes because all but one of its identifying TRs were recorded during the same data collection session and were from the same species. The CZs identified along the eastbound lanes are not strongly aligned with CZs along the westbound lanes. The Relationship of First Order Patterns to Landscape-Scale Features At the landscape scale the different densities of TRs in each of the paired sub-areas was correlated to significant differences in the composition of the cover type, slope, and aspect classes of the surrounding landscape. The ESR of contiguous patches with a consistent slope was also significantly different between the sub-areas at all three study sites, but not the average patch size. The results are described in detail below. 86

PAGE 102

N r I-70 Figure 4.8 Location of crossing TRs that make up CZs in the sub-area adjacent to CMRatVPS. 87

PAGE 103

Eagle County Summit County N i Figure 4.9 Location of crossing TRs that make up CZs in the sub-area not adjacent to C1\1R at VPS 88

PAGE 104

Composition I examined and compared the composition of the cover type, slope, and aspect in the landscape surrounding the pairs of study site sub-areas. Because of the large scale of this analysis I lumped all structurally similar cover types together, creating three cover classes: forest, shrub, and grass/forb. At the landscape scale, animals are most likely to choose habitat based on general structural considerations (forested cover vs. open grassy areas) rather than specific floristic characteristics (mixed forest vs. pure coniferous forest). The composition of the cover type, slope, and aspect of the landscape surrounding the highway differed significantly between each of the sub-area pairs. At all three sites, sub-areas with higher crossing densities were dominated by woody vegetation (Table 4.5). However, although TCP south had more forest cover then TCP north, at both VP and VPS the sub-areas with fewer crossings had the higher proportion of forest cover. At both VP and VPS the landscape surrounding the more frequently crossed sub-area wa8 significantly less steep, but the opposite was true at TCP (Table 4.5). The dominant aspect of the landscape surrounding sub-areas with higher crossing rates was different at each study site. The sub-areas with lower crossing rates were comprised predominantly by northeastern, southwestern, and western facing slopes (Table 4.5).-89

PAGE 105

Table 4.5 Comparisons of sub-area pairs by composition, p-values indicate the likelihood that pairs are different from each other by chance Dominant Cover Type Class Dominant SloEe Class{es) Dominant AsEect Class(es) TCP South Forest+ shrub (36 + 36 %) 11-30 (62 %) SE-NW(65 %) North Grassland (52%) 0-10 (63 %) NINE,IE, SW/W (73%) p=O.OOO p=O.OOO p =0.000 VP East Forest (63 %) 10-25 (80 %) NS (74%) West Forest (74 %) 15-30 (79 %) NINE, SW/W (78 %) p=O.OOO p=O.OOO p=O.OO VPS CMR. Forest (65 %) 6-15 (55%) NW/NINE, S/SW (66 %) NotCMR. Forest (74 %) 16-25 (53%) NEIE, SW IW (70 %) =0.000 p=O.OOO p= 0.000 Note: the values in parentheses indicate the percent of the landscape comprised of the dominant class(es); the reported p-values are derived from chi-square test comparing the distributions of cover classes of each sub-area pair. Complexity The complexity of cover type, slope, and aspect classes in the landscape surrounding each sub-area also varied between each of the sub-area pairs. However, no consistent trend in this variation was correlated with sub-areas that experienced higher rates of crossing. At TCP, animals crossed the road more often in the southern portion of the study area. Visually, the two sub-areas of the study site appeared very different. A few, large, contiguous patches of vegetation dominated the roadside views in the north end where as views in the south end encompassed many medium-sized, interdigitated blocks (See Figure 3.3). However, quantitative comparisons of the twQ sub-areas provided only moderate support for these observed patterns. Although on average TCP south had smaller, more complex habitat patches, the difference in size was not significant and the difference in complexity was only marginally significant (Table 4.6). 90

PAGE 106

Table 4.6 Comparison oflandscapes metrics associated with first order patterns. The likelihood that the paired values are different from each other by chance is indicated by the p-values TCP South North VP East West VPS Cover Slope Aspect Mean Patch Area to Mean Patch Patch Area to Mean Patch Patch Edge Ratio* Size (ha) Edge Ratio* Size (ha) Size(ha) 92.4 10.76 12.4 2.53 6.4 170.4 11.29 26.2 2.68 7.5 p=0.25 p=0.078 p=0;344 p=O.OO p=0.485 1701.1 37.53 38.8 4.72 79.7 1632.1 34.52 16.5 4.08 70.9 p= p=0.424 p=O.OO p=O.OO p=0.534 0.831 Patch Area to Edge Ratio 2.44 2.27 p=O.OO 6.00 5.60 p=0.68 CMR 1318.6 33.17 31.7 4.45 73.3 5.97 Not 1712.5 37.29 22.4 4.26 76.5 5.76 __MR. p = 0.32 p = 0.21 p = 0.58 p = 0.00 p = 0.84 p = 0.33 Note: The reported p-values are derived from an independent samples t-test comparing the paired values The larger the value, the less complex the shape of the patch At TCP,. the patterns created by topography also differed between the two study site sub-areas. The north of the study site, which had fewer crossing TRs, appeared less complex topographically than the south end. With a very large, flat area west of the roadway and smooth slopes to the east, vistas were wide and unbroken. Conversely, the south end of the study area appeared rough, with fewer, smaller flat areas as well as many abrupt changes in slope and aspect. Quantitative analysis of the 'distribution, shape, and size of patches created by changes in slope or aspect reflected these visual impressions. On average, contiguous areas with a consistent slope or aspect were smaller and their complexity was significantly greater at TCP south (Table 4.6). However, aspect complexity was significantly greater at TCP north. This result is counterintuitive, because in the north end, the average patch size is greater and the average edge length is shorter as compared with the south end. The average ESR is apparently skewed by a preponderance of small patches with complex shapes in the northern half of the study area. 91

PAGE 107

At VP east animals crossed the road most often than VP west, but this first order pattern of crossing TR density was not associated with any difference in cover-type complexity. The patterns created by patches of different cover-types at VP are visually and quantitatively similar on both sides of the Pass. Mediumto large sized patches of vegetation that were well interdigitated dominated throughout the study site (See Figure 3.6). Neither the size of these patches, nor the complexity of their shapes was significantly different between the two sides (Table 4.6). Variations in topography were, however, correlated with the first order TR patterns at VP. Visually, VP west offers strikingly steeper slopes than VP east, and there is a restricted, canyon-like feel to many of the views from the roadside. On the east side slopes are gentler and vistas are wide and unbroken. Quantitative analysis of the topography reflected these visual impressions. On average, VP east had larger contiguous areas with a consistent slope or aspect as compared with VP west, and complexity ofthese slope and aspect patches was also greater on the west side. These differences were highly significant for slope (Table 4.6). At VPS, animals crossed the roads more frequently near CMR, but this first order pattern of TRs was not reflected by differences in cover type complexity at this study site. The patterns created by cover type at VPS CMR and throughout the rest of the study area were similar both visually and quantitatively. Well interdigitated mediumto large-sized patches of vegetation dominate both sub-areas. Neither the size of these patches, nor the complexity of their shapes was significantly different between the sub-areas (Table 4.6). Variations in topography between the two sub areas at VPS were inconsistent. Visually, VPS CMR looks very similar to the rest of the east side of the Pass, and as a consequence, very different from the west side of the Pass. Adjacent to CMR, there was no difference in the size or shape of contiguous areas of a consistent aspect quantitatively, as compared with the rest of VPS. However, contiguous areas of a consistent slope were larger in the sub-area 92

PAGE 108

adjacent to CMR and were significantly less complex in shape as compared with the rest of the study site {Table 4.6). Other Influences from the Landscape-Scale At TCP, the patterns of cover and topography were the major difference between the two sub-areas of the study site. At VP and VPS, however, additional variations between the sub-areas existed, which could influence where and how often animals crossed the highway. At VP, differences in the design of the highway between the two sub-areas affected crossing rates, and at VPS the presence of CMR had a strong influence on crossing rates. The placement of underpasses within VP east as compared with VP west crossing rates, as demonstrated by comparing the number of crossing TRslkm of roadway with the amount of underpass available in that km of roadway. Animals crossed at-grade more often where there were fewer meters of underpass available. On the west side of the Pass there are more underpasses, and the alignment of the underpasses in the westand eastbound lanes allows animals to cross under both lanes directly. There were fewer underpasses on the east side of the and the five underpasses in the eastbound lanes were not mirrored by underpasses in the westbound lanes. Animals that pass through an underpass in the eastbound lanes were forced to cross the westbound lanes at -grade if they want to continue on in the same direction. There was a clear spike in crossing rates along the westbound lanes, related to where the underpasses are located along the eastbound lanes, on the east side ofthe Pass (Figure 4.1 0). At VPS, the location and frequency of animals crossing at-grade was strongly correlated to CMR (Figure 4.3). 93

PAGE 109

The Relationship of Second Order Patterns to Local Scale Features In general, the roadside variables measured directly in the field were not significantly correlated with CZs, and those that were had inconsistent relationships with CZs across the three study sites. The roadside variables measured with Arc View from digital data layers created with either field measurements, or from remotely acquired data, were more likely to have significant relationships with CZ locations. However, only some of those relationships were consistent across the three study sites. Variables from the roadside that were significantly correlated to CZ location included the aspect, cover type, distance to the nearest drainage, and the distance to the forest boundary. Additionally, CZs were not located portions of the road that were obstructed by barriers such as Jersey barriers, guardrails, or cliffs. The relationship of CZ locations to all the measured local-scale variables is described in detail below. Features Measured Directly in the Field At all three stUdy sites, none of the features measured in the field were significantly different between CZ points and random points across all study sites. Only one, LOSTRb was significantly different at two sites, but the relationship at TCP was the opposite of the relationship at VP. A complete listing of results of habitat characteristics measured in the field appears in Table 4.7. 94

PAGE 110

Table 4.7 Average values of measurements taken at CZs and random point locations TCP VP VPS 1.6 0.8 1.6 0.8 1.6 0.8 nwv cz 52.0 50.4 55.0 60.5 51.7 50.0 Random 54.6 47.6 58.4 59.4 58.1 56.1 DWV20 cz 34.0 22.2 46.6 54.2 38.9 36.2 Random 31.5 27.4 32.2 37.6 36.3 32.1 LOSAR* cz 293.8 289.9 268.8 261.9 189.3 Random 287.7 273.6 259.1 282.2 262.3 LOSTRa* cz 0.82 0.73 1.57 1.46 1.50 1.38 Random 1.03 0.90 1.21 1.07 1.23 1.07 LOSTRb cz 0.56 0.70 1.48 1.30 LIS 1.00 Random 0.81 0.88 1.05 0.87 1.09 0.93 DTSR cz 185.2 300.0 283.9 Random 224.5 295.86 295.9 DTHA* cz 425.9 464.0 408.4 Random 467.4 451.4 443.8 MINSLu cz 16.8 4.9 2.96 Random 14.4 9.6 9.11 MAXSL*** cz 9.4 20.4 20.9 Random 18.6 19.0 19.9 Note: See Chapter 3 for a complete description of each measurement; pairs which are significantly different from each other at a = 0.05 appear in bold. Values given are the diStance in meters Values given indicate the level of visual obstruction where 0.00 =completely obstructed, 2.0 =no obstruction Values given are degrees ofslope At TCP, I measured local-scale habitat characteristics at 44 points within CZs and at 89 random points. Four of the measured characteristics differed significantly between CZs and random points. The maximum slope was smaller at CZ points, the distance to a node of human activity was smaller at CZ points, and the peripheral view of the roadway at 20 m from the roadside was more obstructed at CZ points for both mediumand large-sized 8llimals (Table 4. 7). 95

PAGE 111

At VP, I measured local-scale habitat characteristics at 28 points within CZs and at 43 random points. Four of the measured characteristics differed significantly between CZ points and random points. The line-of-sight down the roadway was shorter at CZ points for medium-sized animals, the distance to a side road was shorter at CZ points, and the distance to woody cover from the roadside was shorter at CZ points for both medium, and large sized animals (Table 4.7). At VPS, I measured local-scale habitat characteristics at 24 points within CZs and at 44 random points. Four of the measured characteristics differed significantly between CZ and random points. Twenty meters from the roadside woody cover was closer to random points from the perspective of both large and medium sized animals, and the peripheral view of the roadway 20 m from the roadside was less obstructed at CZ points for both medium and large sized animals (Table 4. 7). Features Measured from GIS Data Layers The relationships ofCZs to cover, slope, and aspect within 100m of the roadside, the forest boundary, drainages, and roadside barriers were measured from GIS data layers. The results of comparisons between the used and the available cover, slope and aspect classes are reported in Table 4.8. The cover type classes associated with CZs at four of the six sub-areas were different from what was available along the entire roadside, but the relationship was inconsistent. At VP west and TCP north, CZs were associated with forest cover, at VP east they were associated with grass cover types, and at VPS CMR the CZs were negatively associated with shrubby cover types. The distribution of slope classes associated with CZs did not differ from what was available along the roadside except for VPS Not CMR (Table 4.8). At this sub-area, 96

PAGE 112

the CZs were positively associated with slopes of 15 degrees or less. Conversely, aspects at CZs differed compared with what was available along the roadside for all sub-areas but TCP south (Table 4.8). However, across these five sub-areas, there was no consistency in the aspect classes that were either positively or negatively correlated with CZs. Table 4.8 Results of chi-square tests comparing the distribution of cover, slope, and aspect classes within 100 m of the CZs to what is available throughout the highway corridor within 100 m of the roadside Study Area Sub-area Cover Type Slope Aspect TCP South End l = 1.02, p > 0.25 i= 13.73, p > 0.25 i = 7 .84, p > 0.25 North End l= 8.39, p < 0.05 i= l0.5l,p>0.25 l = 76.94, p < 0.001 VP East Side l= 33.04, p <0.001 r!= 3.51, p > o.25 2_ X -20.01, p < 0.05 WestSide x2= 25.77, P < o.oo1 l= 6.70, p > 0.25 x2= 15.43, p < o.to VPS CMR x2= 7.92, p <0.05 x2= 5.36, p > 0.25 l = 24.49, p < 0.001 NotCMR x2 = o.92, p > o.25 x2= 40.24, p < o.oo1 x2=45.75, p < 0.001 The results of the comparisons between the actual and expected distribution of crossing TRs in relationship to the locations of barriers, drainages, and the forest boundary are reported in tables 4.9, 4.10 and 4.11, respectively. No results are reported for the distance to barriers for TCP north, because the scripts used to generate the expected values could not run a Monte Carlo simulation with the small number of points required in that sub-area. A visual analysis of all sub-areas suggests that crossing zones were associated with the ends of roadside barriers (e.g., figures 4.10, 4.11, 4.12, 4.13). However, a quantitative assessment of this relationship produced inconsistent results. Crossing 97

PAGE 113

\0 00 1-70 N i UDderpass Barriers 6 TRs Figure 4.10 Detail of the summer crossing TRs that made up CZs, in relationship to barriers and underpasses, on the east side of Vail Pass.

PAGE 114

\0 \0 ---=-1-70 N -Underpass i -Barriers ll. NotCMR.lRs (l) CMR 1Rs / Figure 4.11 Detail ofthe winter crossing TRs that made up CZs in relationship to barriers and underpasses

PAGE 115

--1-70 -Barriers 6 TRs N i \ SummitCollllly ,...... 0 0 Figure 4.12 Detail of summer crossing TRs that made up CZs, in relationship to barriers, on the west side of Vail Pass

PAGE 116

,_. 0 ,_. 1-70 Barriers 1:>. TRs N i Eaalo CoiUll)' Summit Couoty Figure 4.13 A detail of winter crossing TRs that made up CZs, in relationship to barriers, on the west side of Vail Pass.

PAGE 117

TRs were significantly closer than expected to the end points of barriers at TCP south, and VP east, but not at the remaining sub-areas. However, significantly fewer crossing TRs than expected were located along the roadside in the middle of barriers at all study sites (Table 4.9). In other words, animals that encountered a barrier crossed the roadway at the barrier's end. They did not wander between the barrier and the roadside, before crossing. Table 4.9 Actual distances compared with expected: distances ofCZs to banier ends and the results of the chi-square test comparing the actual with the expected number of TRS located mid-barrier Study Actual Expected t-test Results Mid-Barrier Chi-Square Area Sub-area Mean Mean Results Distance Distance (m) (m TCP South 43.1 52.2 t= 1.63, p <0.10 x2 = 8.39, p <:: 0.01 End North 618.8 End VP East Side 202.31 259.64 t=2.54, p < x2= 8.67, p < o.o1 0.001 West 232.48 201.99 t = 0.94, p > 0.25 x2= 10.57, p < 0.01 Side VPS CMR 286.75 302.87 t = 0.67, p > 0.25 x2 = 13.26, p < 0.01 NotCMR 208.8 196.78 t = 0.48, p > 0.25 x2= 8.39, p < o.o1 p-values indicate the likelihood that the actual number of points located mid-barrier is different from the expected number by chance. The relationship of crossing TRs to the nearest drainage intersecting the highway was inconsistent across the study areas. Crossing TRs were significantly closer than expected to drainages at TCP south, VP east, and VPS not CMR. Crossing TRs were significantly farther from drainages than expected at TCP north and VP west. The distance to drainages did not differ from what was expected at VPS CMR (Table 4.1 0). 102

PAGE 118

Table 4.10 Actual distances compared with expected distances of CZs to the nearest drainage that intersects the road Study Area Sub-area Actual Mean Expected Mean t-test Results Distance (m) Distance (m) TCP South End 216.58 447.09 t = 3.509, p < 0.001 North End 3402.85 2299.17 t = 3.650, p < 0.001 VP East Side 188.53 463.77 t = 6.89, p < 0.001 WestSide 3076.09 1583.11 t = 0.84, p < 0.25 VPS CMR 666.78 601.10 t = 1.14, p < 0.10 NotCMR 654.38 1258.14 t = 3.7o, e < o.oo1 The relationship of crossing TRs to the forest edge was also inconsistent across study areas. Crossing TRs were significantly closer than expected to the forest edge throughout TCP, marginally so at VP east, but not at the remaining sub-areas {Table 4.11 ). Table 4.11 Relationship of CZs to the forest edge Study Area Sub-area Actual Mean Expected Mean t-test results Distance (m) Distance (m) TCP South End 20.72 26.43 t = 2.17, p < 0.05 North End 26.15 68.54 t = 3.84, p < 0.001 VP East Side 49.93 58.60 t = 2.62, p < 0.10 WestSide 43.66 41.44 t = 0.34, p > 0.25 VPS CMR 57.66 54.39 t = 0.15, p > 0.25 NotCMR 35.78 47.06 t = 2.14, p = <0.01 Underpass Use Animals used underpaSses to cross under the highway at all three study sites even though roadways at the study sites were unfenced and they were free to cross at grade. Due to the superior tracking medium placed in the underpasses, it was possible to record a greater range of species using the underpasses than crossing at grade. Animals used the underpasses throughout the year. Although the type of data 103

PAGE 119

that were collected cannot be used to indicate a clear preference for certain underpass designs, they do demonstrate that a variety of designs is acceptable to a range of species. Trout Creek Pass The characteristics of the structures monitored for use at TCP, the number of times at least one set of tracks was observed passing through, and the number oftimes at least one animal was recorded crossing at grade next to an underpass (an end-run) are reported in Table 4.12. I .checked I 0 underpasses a total of 482 times, and at least one set of tracks was recorded passing through 23 % of the time. At least one set of tracks indicating an end-run was recorded 42 % of the time. All of the recorded end-runs were made by deer. The ratio of each underpass' use (number of times used/number of times checked) based on the total number of through-passes, by species, is reported in Table 4.13. Because tracks from more than one species were often recorded when an underpass was checked for tracks, this ratio exceeds 1.0 for some underpasses. Multiple track sets from one species, especially deer, rabbits, and coyotes were commonly recorded. In most cases there was no way of knowing if these occasions represented one animal crossing many times or many animals crossing once. Track sets from two or more species were also relatively common, and consisted of both predator and prey species in many cases. In general, underpass use by all species remained consistent throughout the year, although mountain lions were not recorded during the summer months (June-August). The four underpasses that I. checked most regularly (MP 215.0, 216.48, 218.4, and 219.2) were used throughout the year. Notable crossing events include a beaver dragging branches through the 104

PAGE 120

MP 219.2 structure on four occasions, and deer beginning to use the concrete bottomed structure at MP 216.15 in late November, 2000, then continuing to use it consistently throughout the remainder of the study. Table 4.12 Characteristics of monitored underpasses at TCP and the number of times at least one through-pass by at least one medium-or large-sized animal was recorded chamber No. No. Through No. of MP Type* Height Widthffotal Length Times Passes(%) End-runs Location (m} Width(m} (m} Checked** {%} 215.0 Single Span 14.0 24 11.50 49 45 (91.8 %) 216.1 3 Chamber 3.0 3.40/10.20 14.00 23 11 ll (47.8 %) (47.8%) 216.5 3 Chamber 3.0 3.05/9.15 18.25 21/85 23 15 (27.1 %) (71.4 %) 216.8 2Chamber 3.4 2.48/4.96 32.70 16 8 12 (50.0 %) (75.0 %) 217.1 1 Chamber 2.5 2.50 27.50 19 1 5 (5.3 %) (26.3 %) 218.0 1 Chamber 2.5 2.50 22.10 17 5 8 (29.4%) (47.1 %) 218.4 3 Chamber 2.9 3.10/9.30 21.30 271105 28 8 (26.7%) (29.6 %) 219.2 3Chamber 3.0 3.10/9.30 21.30 29/105 31 18 (29.5 %) (62.1 %) 221.9 3Chamber 2.4 3.10/9.30 14.60 29 4 7 (13.6 %) (24.1 %) 222.60 Single Span 3.1 7.34 1l.l0 34 1 6 (2.9 (17.6 %) Single span bridges have natural floors; all other structures have concrete floors. **For 216.48, 218.4, and 219.2 the first number in this column is the number oftimes the road surrounding the culvert was checked for evidence of end-running. The second number is the total number of time the trackbeds were checked for tracks. The at-grade road side at 215.0 was never checked for end-runs. Table 4.13 Number of times at least one individual of a species used each underpass at TCP 215.0 216.1 216.5 261.8 217.1 218.0 218.4 219.2 221.9 222.6 s cies Deer 39 4 5 Elk Coyote 10 2 9 5 2 7 8 4 Fox 7 1 5 15 Bobcat 1 3 1 9 6 Mt. Lion -2 1 3 2 3 Rabbit ll 12 20 5 5 16 49 Weasel 1 1 2 8 4 1 Total 62 24 47 13 6 6 48 82 5 Crossing Ratio 1.26 1.04 0.55 0.81 0.31 0.35 0.46 0.78 0.17 0.03 105

PAGE 121

Rabbits, deer, and coyotes were the species most commonly recorded using underpasses. Coyotes and rabbits used the greatest number of different underpasses, but deer used their favored underpass (MP 215.0) most consistently. The single span bridge at MP 215.0 had a natural floor, was very open, and received the most consistent levels of use, including large numbers of deer as well as some bobcats and coyotes. However, although larger structures were generally used more then smaller ones, size did not guarantee use. The single span bridge and the three-chambered culvert located in the north end of the study area received no or very low rates of use. Vail Pass The characteristics of the structures monitored for use at VP, the number of times at least one set of tracks was observed passing through, and the number of end-runs are reported in Table 4.14. I checked four underpasses a total of347 times, and at least one set of tracks was recorded 91% of the time. At least one set of tracks end running an underpass was found 29% of the time. Table 4.15 reports the number of observations by species, when at least one track set was recorded. Each underpass' ratio of crossing, based on the total numberofthrough passes, by species, and the number of times each underpass was checked, is also reported. 106

PAGE 122

Table 4.14 Characteristics of the underpasses monitored for use at VP, including the number of times checked for tracks, and used by at least one medium-or large-sized animal to cross through Location Type Height Length Width No. Times At Least One At Least (m)* (m) (m) Checked** Through Pass One End-run MP 183.0 Two adjacent 26.0 2-lane bridges 3.9 (both 20.9 90 86 (95.5 o/o) 24 (26.7 %) spans) MP 184.9 Two adjacent 26.0 2-lane bridges 13.4 (both 128.0 84 81 (96.4 %) 16 (19.1 %) spans) MP 190.8 One 2-lane bridge 5.6 12.9 45.0 88 75 (85.2 %) 38 (31.8 %) MP 191.8 One 2-Iane bridge 10.8 12.9 71.1 85 75 (88.2 %) 25 (29.4 %) **The height of most bridges varied with topography; the maximum height is reported. ***It was not possible to access trackbeds due to highway maintenance and repair work on some occasions. Table 4.15 Number of times at least one individual used the underpasses at VP Species Underpass MP 183.0 MP 184.9 MP 190.8 MP 191.8 Deer 86 81 75 75 Elk 3 6 Moose 1 1 Coyote 10 15 1 5 Fox 3 1 6 Mountain. Lion 1 1 2 Bear 2 Totals 102 107 83 83 Crossing Ratio 1.13 1.27. 0.94 0.98 Deer were the species most commonly recorded using all underpasses, and they used the underpasses on the west side of the Pass heavily and consistently. During June, July, and August of both years there were often so many deer tracks in the trackbeds ofMP 183.0 and 184.9 that they obliterated one another and I could not count them accurately. Use was high throughout.the summer, and then dropped off in the fall, reflecting the shift of deer and elk to lower elevations for the winter months. The less commonly recorded species were most likely to be recorded in June, September, and October, perhaps also reflecting seasonal migrations which 107

PAGE 123

were more likely to bring the animals into contact with the road. Coyote tracks were recorded more commonly in early summer and fall both below the roadway and at grade, and the two sets of bear tracks were recorded at ahnost the exact same location under 184.9 in October 2000 and 2001, just prior to hibernation. Vail Pass Snow The characteristics of the structures monitored for use, the number of times at least one trail was recorded passing through, and the total number of trails observed are reported in Table 4.16. I checked seven underpasses a total of 108 times, and at least one set of tracks was recorded 49 % of the time. Only one end-run was recorded, by a marten at MP 190.8. Because the density oftrails going through the underpasses was low during winter and snow cover provided an excellent tracking medium, it was possible to count all tracks sets with great accuracy. However, when multiple trails from the same species were observed traveling though an underpass, it was still not possible to determine if this represented one animal passing through multiple times or multiple animals passing through once. Table 4.17 reports the number of observations when at least one track set was recorded, by species, and Table 4.18 reports the total number of trails that were recorded. Each underpass' ratio of crossing, based on the total number of through passes, by species, and the number of times each underpass was checked, is also reported. The data indicate that coyotes used the greatest variety of underpass most consistently and MP 192.5 and 184.9 were the most commonly used underpass in winter. Additionally, the rate of animal passage through underpasses was higher during March, as compared with the December-February period. 108

PAGE 124

Table 4.16 Characteristics of the underpasses monitored for use at VPS, and the number of times it was used by at least one medium-or largesized animal to cross through At Least One Total Location Type Height* Length Width No. Times Through Pass Number (m} {mj (m) Checked of Trails MP Two adjacent 2183.0 lane bridges 3.9 26.0 21.9 17 5 (29.4 %) 5 MP. Two adjacent 2184.5 lane bridges 13.5 12.9 218.0 13 10 (76.9 %) 24 MP Two adjacent 2184.9 lane bridges 13.4 12.9 128.0 19 13 (68.4 %) 30 MP One2-lane 190.8 bridge(eb) 5.6 12.9 45.0 17 7 (41.2 %) 13 MP One2-lane 191.4 bridge(eb) 13.2 12.9 13 5 (38.5 %) 7 MP One2-lane 191.8 bridge(eb) 10.8 12.9 71.1 17 6(35.3 %) 11 MP One 2-Iane 192.5 bridge (eb) 10.5 13.0 103.5 12 7 (58.3 %) 32 *The height of most bridges varied with topography; the maximum height is reported. Table 4.17 Number of times at least one individual used the underpasses at VPS Species MP 183.0 MP 184.5 MP 184.9 MP 190.8 MP 191.4 MP 191.8 MP 192.5 Coyote 2 6 7. 9 3 4 8 Weasel I 4 1 Marten 1 7 8 6 Hare 1 4 1 9 Elk 2 I Totals 5 15 24 10 3 5 23 Crossing Ratio 0.29 1.15 1.26 0.59 0.23 0.29 1.92 Table 4.18 The total number of animals that used the underpasses at VPS Unde!J!asS Species MP I83.0 MP 184.5 MP 184.9 MP I90.8 MP 191.4 MP 191.8 MP 192.5 Coyote 2 14 12 12 5 10 15 Weasel 1 1 3 1 Marten 1 6 9 5 Hare I I 12 Elk 2 4 Totals 5 23 29 13 5 11 32 109

PAGE 125

Because I could conduct tracking away from the roadside only when there was snow on the ground, my field results estimate the relative distribution of different species in my study areas only for the winter months. The Colorado Division of Wildlife (CDOW) estimates deer and elk populations after the hunting season each year throughout the state. Trout Creek Pass In the winter of 2000/200 I I conducted track counts along I 0 transects away from the road, then conducted counts along four more transects during the winter of 200112002 for a total of I4 transects sampled. The transects were evenly split between the north end and the south end of the study area, and at least one transect was .located in each of the major cover types. More tracks were recorded in the south end (368 vs. 272) but only because of the much higher number of deer tracks observed in the south end ( 40.3/transect vs. 8.9/transect). Other common species included coyotes, elk, and cottontails, all ofwhich were more numerous in the north end of the study area (Table 4.I9). The Colorado Division of Wildlife (CDOW) splits the state into Deer J\nalysis Units (DJ\Us) and Elk J\nalysis Units (EJ\Us) based on areas that have similar habitat and animal densities. The entire TCP study site is located within the same DJ\U and EJ\U. Because all ofTCP is located within a single analysis unit for both species, elk and mule deer densities and habitat quality should be relatively uniform throughout the study site. The CDOW considers the habitat in the Trout Creek Pass area to be high quality for elk, but not for deer. 110

PAGE 126

Table 4.19 The nwnber of animal trails/transect recorded in the snow along off-road transects at TCP and VPS, by species TCP North End* South End** Deer 8.9 (62) 40.3 (282) Elk 14.6 (102) 0.2 (2) Coyote 3.6 (25) 2.7 (19) Bobcat 0.1 (1) 1.1 (8) Mountain Lion 0.1 (1) VPS East*** 4.5 (18) 1.0 (4) Weasel species 0.1 (1) 0.1 (I) 1.2 (5) American Marten 1.5 (6) Rabbits/Hares 11.1 (78) 7.3 (51) 24.0 (96) Total 38.9(272) 52.6(368). 31.7(131) Seven transects sampled, total number of animals given in parentheses. ** Seven transects sampled, total nUmber of animals given in parentheses. Five transects sampled, total number of animals given in parentheses. One transect sampled, total number of animals given in parentheses. Vail Pass VPS West**** 4.0 (4) 1.0(1) 36.0 (36) 41.0 (41) The CDOW population estimates offer some perspective on the summer time densities of deer and elk on Vail Pass. Vail Pass and its associated ridge line split the area into two separate DAUs. The estimated density of deer is 4.2 deerlha east of the Pass and 5.9 deer/hawest of the Pass. The Vail Pass area is split into three EAUs, including the eastern side of the Pass, and 1-70 serves as the border between two additional units on the western side of the Pass. The estimated densities of elk in these three units are 3.5, 4.6, and 3.5 elk/ha, respectively. The CDOW does not designate any of the DAUs or EAUs located in the Vail Pass area as quality habitat. Vail Pass Snow In the winter of2000/2001, I conducted track counts along three transects away from the road, then conducted counts along two more transects during the winter of 2001/2002 for a total of five transects sampled. Topography and avalanche danger constrained the locations where transects could be safely conducted on the west side ofthe Pass, and only one of the transects was located on that side. The 111

PAGE 127

remaining four transects were evenly split between the north side and south side of the highway on the east side of the Pass. The overall density of animals at VPS was low (Table 4.19). An average of34 animal trails crossed each transect, and the most common species recorded was snowshoe hare (76 %). Other species recorded, in order of prevalence, included elk, coyote, weasel species, marten, and red fox. 112

PAGE 128

CHAPTER VARIABLES THAT IDENTIFY WILDLIFE HIGHWAY CROSSING ZONES: DISCUSSION Introduction My results indicate that midand large sized mammals do not cross highways at random. Instead, they focus their activity at both the landscape and the local scale. At the local scale, crossing activity becomes sufficiently concentrated to create definable crossingzones (CZs) that are on average, about 500 min length. Landscape features were linked to the large-scale activity zones. However, most of the variables that I tested for association with CZs did not yield consistently significant relationships. Further, the variables that were significantly associated with CZs had often had inconsistent relationships across the three study areas. This result is not surprising. Like niost landscapes, Trout Creek Pass, Vail Pass, and Vail Pass Snow are to a great degree unique and do not represent replicate samples. Variables that may regulate animal movement, including structure of the surrounding landscape, highway footprint, traffic volumes, and snow depths varied substantially between the three data sets that I collected. Although it may have been to average the measured variables across these three landscapes and use the results to predict the most likely location for a CZ, I do not believe the outcome would be biologically relevant. As noted above, the variables that I measured at each site combined to form unique conditions, and should be treated as such. However, it is still possible to examine the data sets for trends, and to consider how the differences of the study areas shaped the results. 113

PAGE 129

The' combination of roadway design and the surrounding landscape, make every highway project is likewise unique. Thus, this qualitative approach should provide insight to highway project planners considering the problem of reducing wildlife/highway conflicts at a specific location. Descriptive Summary of Tracks and Their Distribution The descriptive summaries of the track records (TRs) indicate that the most common species present along the roadside, the overall rate of crossing, and the propensity of a species to cross differed among the three study sites (tables 4.1, 4.2, 4.3). These differences were linked to habitat characteristics at the three study sites. At both TCP and VP, the habitat was good quality for deer and elk, the most commonly recorded species. At VPS, deep snows made the area inhospitable for deer and elk, and may have served to draw coyotes to the roadside, especially near C:MR, where travel through the winter environment was easier and extra foraging opportunities were present. Snow also appeared to be a strong influence the crossing rate as well. Although the crossing rate at TCP did not differ significantly from VP or VPS, VP and VPS did differ significantly from each other. Additionally, coyotes at VPS were the only species that crossed the roadway more often then not (74.5% of time that they approached the roadway; t5ables 4.1, 4.2, 4.3). They may have been the most likely of all the species/location combinations to cross at-grade because they have no special physiological adaptations to the severe winter conditions (deep snows, cold temperatures) in the Vail Pass area, and consequently are on a tight energy budget during winter. Coyotes at VPS probably only approached the roadside when they were intent on crossing. 114

PAGE 130

First Order Patterns and Landscape-Scale Variables The first order patterns of TR distribution were correlated with variables from the landscapes surrounding the highways at all three study sites. However, although most of these differences were statistically significant, they must be examined thoughtfully to determine if they represent biologically significance. Composition and Complexity At all three sites, the sub-areas with higher densities of TRs in general, and crossing TRs specifically, corresponded to landscape composition and. to a lesser degree, complexity (Tables 4.5, 4.6). At TCP, significantly more TRs were located in the south end even though CDOW population density estimates indicate consitatent deer and elk populations throughout this study area. At VP, more TRs than expected were recorded at VP east even though CDOW estimates that both deer and elk populations are lower there, as compared with VP west At VPS, significantly more TRs were recorded near CMR even though my off-road track transects suggest that distribution of predators, the species that crossed the road most often when there was snow on the ground, should be relatively even throughout VPS. All high density TR sub-areas occurred in landscapes with at least 70 % woody cover, as would be expected when dealing with species such as deer, elk and coyote that prefer habitats with some woody cover. This result supports previous research that indicates animal activity at the roadside is highest adjacent to forest cover (Lyon, 1979;Rost and Baily, 1979; Finder et al., 1999). However, the complexity of cover-types patches did not appear to play a role, as the ESR of patches differed 115

PAGE 131

significantly only at TCP, and only marginally so. This finding supports neither the theoretical work of Stamps et al. ( 1987), which indicates movement rates should be higher in complex landscapes, or the empirical results of Hubbard (2000) and Haas (2000), both of which indicate that crossing is more likely adjacent to large, homogeneous blocks ofhabitat. Both slope composition and slope complexity were linked to TR density. At VP and VPS, the higher densities ofTRs were associated with significantly slopes and lower ESR. This makes sense, as landscapes comprised of shallow, consistent slopes are easier to traverse, and animals should therefore be more likely to use these landscapes and come into contact with the roadside. These results mirror the fmdings of Alexander and Waters (2000). At TCP, the opposite relationship to slope composition and complexity was statistically significant. However, although TCP south had more. then twice as much area over 20" in slope as compared to TCP north, 76% ofTCP south was still under 20", providing plenty of easy travel terrain. Meanwhile, TCP south had half again as much shrub cover as TCP north, the preferred forage of deer. Along the roadside, the shrubby areas were well interspersed with grass and forest patches, providing the mix of food and cover that deer prefer (Fitzgerald et al., 1994). At TCP north, a few large blocks of grass, shrub, and forest dominated the landscape. Directly adjacent to the highway, the land cover is mainly grass, and on the west side ofUS 24 the forest edge is about 1.5 km away. Mule deer, the species that most commonly crossed US 24 during the study, generally avoid open areas of that size. The primary role of preferred cover type, and the secondary role of slope in creating landscapes that bring animals to a roadside sense from a biological 116

PAGE 132

point of view, considering the habitat preferences of the species present at my study sites. Aspect, which influences cover-type, average daily temperature, and snow depths in winter, is also known to affect habitat use by many species. Therefore, aspect could reasonably be expected to influence where midand large-sized mammals approach highways. However, the dominant aspects of landscapes surrounding the roadways with higher crossing rates differed at all three study sites. Around the sub-areas with lower crossing rates, northeast-, southwest-, and west facing slopes dominated. However, the lack of a consistent relationship between aspect and TR density is not surprising. The predominant direction of-a roadway that runs in a relatively straight-line direction along a valley bottom (e.g., US 24 and 1-70 in the study site sub-areas) dictates the aspect of the adjacent landscape to a large extent. The orientation of the roadway does not have this same effect on cover-type or slope composition. Thus, even if animals are cueing on prerfered aspects as they choose .habitats surrounding roads, their choice may be swamped by the overall influence of roadway orientation. As a result, only very general differences between the landscapes of low and high rates of use may be apparent, and the difference may have no biological significance. Thus, at the landscape scale, it was impossible to determine if aspect influenced where midand large-sized mammals approached the roadway at my study sites. Other Landscape Scale Influences In addition to cover type and slope, four other variables influenced the first order distribution ofTRs; 1) the number and quality of-underpasses within an entire sub area; 2) the configuration ofthose underpasses; 3) the orientation of landforms that can act as linear guideways; and 4) at VPS, the presence ofCMR. The influence of 117

PAGE 133

these four variables is discussed in detail below. Although I was able to quantitatively assess the influence of underpasses and of CMR on TR distribution, my examination of linear guideways is strictly qualitative because I had no objective means of identifying what constituted a linear feature. At VP there was a strong relationship between number of at-grade crossings and the meters ofunderpasslkm of highway. The number of animals crossing at-grade was higher along the stretches of roadway with fewer meters of underpasseslkm roadway (Figure 4.10). The opposite relationship between rate of at-grade crossing and meters of underpass available occurred at TCP. However, this was expected. At VP, mule deer readily used the underpasses that were present, but at TCP they did not. Instead, the deer followed the drainages to the road, and then crossed at-grade. At VPS, the distribution of TRs was not negatively or positively associated with underpass location. The configuration of underpasses was also important along 1-70. At VP west and VPS Not CMR, only 8.2 and 5.1% of all crossing TRs, respectively, were located along the portion of the highway where underpasses in the eastand west-bound lanes were aligned with one another (MP 183.0MP 185.5; Figure 3.6). In this area, animals could pass directly under the entire roadway, and they were far less likely to make an at-grade crossing. Additionally, on the east side of the Pass the five underpasses in the eastbound lanes were not mirrored by underpasses in the westbound lanes. Animals that passed through an underpass in the eastbound lanes were forced to cross the westbound lanes at-grade ifthey wanted to continue on in the same direction. At VP east there was a clear spike in crossing rates along the westbound lanes, related to where the underpasses are located along the eastbound lanes (Figure 4.1 0). At VPS, the influence of CMR made it difficult to discern the relationship of TR distribution to underpass location in this area. 118

PAGE 134

Wildlife professionals commonly assert that drainages and ridgelines act as guideways for animals moving through the landscape. A visual analysis of mapped TRs at TCP and VP suggests that the first order patterns of TR distribution could have been influenced by the orientation such guideways. At TCP north and VP west, which had lower densities of TRs, most of the predominant ridge lines and drainages ran parallel to the road and did not act to bring animals to the roadside. Additionally, the four major drainages that did bisect the highway at VP west were spanned by large bridges that provided exceptional opportunities for animals to cross under the highway, further reducing the likelihood of at-grade crossing in this sub-area. Conversely, at the south end ofTCP, box culverts (which deer were reluctant) to enter spanned the major drainages that guided animals to the highway. Consequently, the number of at-grade crossings in these locations was high. At VPS, the relationship ofTR distribution to drainages and ridgelines was unclear. Snow depths associated with both types of features vary with the prevailing wind and may have the greatest influence on animal movement in wintertime. At VPS, the location and frequency of animals crossing at-grade was strongly correlated to CMR (68.4% of all crossing occurred adjacent to CMR, Figure 4.3). In winter, snow depths restrict both mobility and food availability for wild animals in the Vail Pass area. This effect is pronounced for coyotes, which are not adapted to moving though deep snows. However, the packed ski trails at CMR and the plowed access roads allow wild animals as well as humans to move through the winter environment with relative ease. Additionally, in winter the availability of trash, birdfeeders, and pet food at CMR may act as a strong attractant to coyotes, the most common species recorded at VPS. 119

PAGE 135

The Relationship of Second Order Patterns to Local Scale Features The relationship ofCZs to local-scale features varied. Most of the features measured directly in field did not differ between CZ and random points. Most of the features measured form GIS data layers did differ statistically between actual use and expected use. However, as with the landscape scale features, it is important to consider if these statistical differences represent biologically significance. Variables Measured Directly in the Field In general, the variables directly in the field were not good indicators of CZ location. At each site, only four of the variables that I measured directly in the field were significantly different between CZ and random points (Table 5.1). The on.ly variable that significant at more than one site was LOSTRb, and its relationship to CZs was inconsistent between the two sites (Table 4.7). At TCP, CZs were associated with more obstructed peripheral lines of sight to the roadway, while at VP they were associated with less obstructed lines of sight. Table 5.1 Results ofthe Mann-Whitney test comparing means oflocal-scale habitat measurements taken at point locations within CZs and random point locations TCP VP VPS 1.6 0.8 1.6 0.8 1.6 0.8 Distance to woody vegetation (DWV) Ns Ns Ns ns DWV20 Ns ns Ns Line of sight at roadside (LOSAR) Ns ns Ns Ns ns Line of sight to roadway (LOSTRa) Ns ns Ns Ns ns Line of sight to roadway LOSTRb ns Ns Distance to side road (DTSR) Ns Dfstance to human activity DTHA Ns ns Minimum slope at roadside (MINSL) Ns Ns ns Maximum slope at roadside (MAXSL) Ns ns Ns Ns Ns Note: See Table 3.1 for a complete description of each measurement; features which differ significantly at a= 0.050 are marked with an . 120

PAGE 136

CZ locations were significantly related to local scale variables measured from both field and remotely generated GIS data layers (discussed below). Therefore, the lack of consistent relationships between the suite of local-scale habitat features measured in the field (Table 5.1) and CZs did not occur because animals do not respond to local-scale habitat variability when crossing highways. Rather, this result is probably related to unique c;onditions at each study site and problems related data collection. A primary problem with the data collected is that it is unlikely that humans see the world the way animals do; therefore, it is probably impossible to re-create a "deer's-eye-view". Consequently, I may have been measuring the wrong variables, and/or my measurements were imprecise. Additionally, my sample sizes may have been inadequate to detect the variation that the a,Il.imals were cueing on. \ I believe imprecise was the most important reason that no consistently significant relationships were apparent. Perception of a location along a roadway is a complex process. Most "locations" are continuous, i.e., they do not have discrete boundaries. Assessment of a location relies on many variables and their interactions. In addition to visual cues, the only sort that I measured, auditory and olfactory cues may be important to wildlife at a local scale. Measuring the total effect of these continuous cues objectively may simply be an unreasonable task. Additionally, in some cases there was insufficient amount of variability in the measured feature to yield a significant difference between CZ and random points, and features' levels of variation differed by site. For example, at TCP, MAXSL was significantly flatter at CZ points compared with random points (Table 4. 7). This relationship was a result of animals avoiding the steep, upgraded roadside 121

PAGE 137

cuts, which were common at TCP south. At both VP and VPS, MAXSL did not vary between the two types of points, but the angle of the roadsides was relatively uniform due to extensive modification of the roadside during construction (FHW A 1978). Consequently, there was little variation in slope to measure and relate to the locations of CZs in the Vail Pass area. Two of the local scale measurements representing human disturbance had relationships with CZ points, DTHA at TCP and DTSR at VPS (Table 4.7). These fmdings bear discussion because, like the MAXSL example, they also illustrate how the unique conditions at a site may drive a relationship, and because they counter the fmdings of other researchers (Clevenger and Waltho 2000; Singleton and Lemkuhl 2000, Gibeau et al., 2002). In both cases, the unique secondary associations of these variables, rather then the variables themselves, were likely the cause.ofthe observed relationships. The association between houses and CZs at TCP occurred because homes were placed along side roads that were located in drainages, and drainages were positively associated with CZs (see below). The topography usually hid the side roads from view, but houses were often built on hillsides and ridgelines. Most of the side roads at VPS were associated with CMR, and the positive association between side roads and CZs was a result of coyotes being attracted to CMR, as discussed above, rather than their attraction to side roads per se. Features Measured fromGIS Data Layers Most of the features measured from GIS data layers were significantly correlated with CZs, regardless of whether the data layers were created from field measurements or from remotely collected data. Additionally, although relationships of CZs to these variables were somewhat inconsistent across the three study sites, 122

PAGE 138

in most cases the differences clearly stemmed from variations in the habitat structure and roadway at the three sites. The comparisons of the cover type, slope, and aspect classes within I 00 m of CZs with the distribution of classes available within I 00 m of the entire roadside indicated that cover type differed at four sub-areas, aspect differed at five, but slope only differed at one {Table-4.8). The mostly non-significant results for slope were not surprising. My experience with the study sites suggests that the 10 m resolution DEMs did not capture the fme-scale variations that animals responded to when choosing a pathway from the adjacent habitat to the roadside. I observed many well-worn roadside game trails that picked a low-angle path through relatively steep slopes. The one significant result was at VPS Not C:MR, where CZs were positively associated with slopes of I5" or less. Snow cover increases the attractiveness of shallow slopes for travel because it obliterates narrow game trails that pick through otherwise inhospitable slopes. Additionally, steep slopes become slippery and/or an avalanche hazard with snow cover. Across the five sub-areas that had a significant relationship with' aspect, there was no consistency in the aspect classes that were either positively or negatively correlated with CZs. As at the landscape-scale, the effect of roadway direction probably swamped .. any effect of animal preference. The four sub-areas with significant relationships to cover also had inconsistent associations with cover type classes, and there was no clear source of biological significance to explain the observed patterns. As with slope, it is likely that variations in cover type to which animals respond to at this scale are too fme-grained to be detected at the resolution of the digital data layers. I23

PAGE 139

The association of CZs with barrier ends appeared strong, despite inconsistencies across the six sub-areas. Crossing zones were positively associated with barrier ends at both at TCP south and VP east (Table 4.9). At VP west there were so many barriers that nearly every location was near to a barrier end. Thus, there was little variability related to barrier ends that could be measured. At VPS, most barriers disappeared under the roadside snowdrifts created by the plows. In essence, these drifts became one continuous barrier along the entire roadway, and once again, there was little variability associated with barrier ends to measure. However, it is important to note that on the west side of Vail Pass, where both median and outer edge barriers are present in many places, all CZs were located along the stretches of road with the fewest barriers, regardless of snow cover (Figures 4.13, 4.14). Animals focused crossing activityon locations with either no barrier at all, or just a median side barrier. Additionally, at all three study sites there were no CZs, and very few crossing TRs, located within the space between a roadside barrier and the roadway. This distribution of crossing TRs differed significantly from expected (Table 4.9). These results suggest that animals do not jump over Jersey barriers or guardrails to enter the roadway, consitent with the results reported by Carbaugh et al. (1975). My results indicate that animals occasionally wander along the roadside, but they apparently do not walk into the narrow space between a barrier and the hardtop before crossing. Previous research indicates that the locations where midand largesized mammals interact with highways are positively associated with drainages (Romin and Bissonette, 1996; Finder et al., 1999; Hubbard et al., 2000), and my results are consistent with these fmdings. The CZs in my study were positively associated with drainages at TCP south, on VP east, and VPS Not CMR (Table 4.1 0). However, 124

PAGE 140

CZs were negatively associated with drainages at TCP north and VP west, and did not have a significant relationship at VPS CMR. These negative associations were not surprising and do not negate a general positive association between CZs and drainages. The CZs at TCP north were strongly associated with the forest edge (see below), which was far away from the few locations where drainages intersected the roadside. At VP west, the large bridges that spanned the drainages created exceptional opportunities for animals to pass under the road. Thus, most animals following these drainages passed under the highway, and only a few at-grade crossing TRs were created near these drainages. Previous research also indicates that the locations where midand large-sized mammals interact with roadways are positively associated with woody cover (Lyon 1979, Rost and Baily, 1979, Rodriguez at al., 1995, Yanes, 1995; Romin and Bissonette, 1996; Finder et al., 1999; Hubbard et al., 2000), and my results support these fmdings as well. CZs were associated with the forest edge in four of the six of the sub-areas, and the relationship was very strong for three of them {TCP north, TCP south, VPS Not CMR; Table 4.10). No significant difference from the expected distribution was apparent at VP west because the forest edge is a consistent distance from the roadside, and consequently, there was no significant variation for animals to cue on. The reason for the non-significant relationship at CMR VPS is probably because the attraction of CMR itself overrode any preference for crossing near cover. However, the relationship of CZs to woody cover bears some additional discussion. Even though my results suggest that midand large-sized mammals are more likely to cross a highway when at least 70 % ofthe surrounding landscape is woody cover {Table and when the cover's edge is no more than about 50 m away (Table 4.11), neither of the amount of woody cover at the roa<;lside, as.measured by a GIS, 125

PAGE 141

nor the nearest individual stand of cover, as measure in the field, was important. This suggests that these species do not need cover right up to the roadside in order to cross. Instead, the most important quality of cover for cover-associated species may simply be that "its there", and that its nearby. Underpass Use Trout Creek Pass Predators, including mountain lions, bobcats, coyotes, and foxes, were recorded passing though structures far more frequently (at least one track set in an underpass on 110 occasions) than they were recorded crossing at-grade (nine crossing individuals out of 45 total recorded at the roadside). This result is probably due, in part, to the superior tracking medium in the structures, but it also suggests that predators do not avoid, and may even prefer these structures when crossing-roads. Conversely, elk were not recorded using any structure, and although deer crossed under the road regularly, they did so only at two structures (MP 215.0 and MP 216.48). These low levels ofuse and the rate at which deer were recorded end running underpasses (Table 4.12) suggest that ungulates at TCP do avoid using underpasses .. For all species, both the characteristics of a structure itself and the surrounding habitat appeared to play a role in the level of use it received. For example, the single span bridge at MP 215.0 had a natural floor and was the most open (height x width/length) structure checked for tracks. It received the most consistent levels of use, including large numbers of deer as well as some bobcats and coyotes. However, a high openness value and a natural floor did not guarantee use. The single span bridge located at the north end of the study area received a very low 126

PAGE 142

rate ofuse (Tables 4.12, 4.13). The cover-type factors that limited at-grade crossing at TCP north probably also limited underpass use. Vail Pass Underpasses atVP were use primarily by deer. Both elk and predators were rarely recorded crossing under the road (Table 4.14). Although predators were also rarely recorded crossing at-grade, elk made 37% of crossing TRs. This suggests that elk avoid using underpasses. The reason for this is unknown, as Clevenger and Waltho (2000) indicate that elk readily used underpasses of similar dimension. At VP the temporal distribution of predator crossings was similar both at underpass and at grade. Predators crossed more often both over and under the roadway during early summer and. fall. This suggests that predators avoid I70 all together during summer, not underpasses in particular. The use ofunderpass.es at VP appears to be most heavily influenced by the pairing of underpasses that allows animals to cross under both the eastbound and the westbound lanes of with ease. The two most heavily used underpasses differed greatly in dimension 4.14) but were both located on the west side of the pass. On this_side pass, the westbound and eastbound alignments are side-by side and the underpasses span the entire highway. On the east side of the pass, the separated by a wide median, and the underpasses available along the eastbound lanes are not mirrored on the westbound lanes (Figure 3.6). The underpasses on the east side of the pass are otherwise similar in construction and dimension to the west side underpasses. Therefore, the ease of crossing the entire highway, rather than an underpass' characteristics appear to play the major role in regulating an underpass' rate of use at VP. 127

PAGE 143

Vail Pass Snow The pattern of underpass use at VPS is similar to the VP pattern. Two of the three most heavily used underpasses were located on the west side of the Pass, as would be expected due to side-by-side configuration of the alignments. However, the underpass that received the highest rate of use was on the east side, located adjacent to CtviR. Because half of the animals that passed through it were coyotes {Table 4.18), CMR appears to play the same role in encouraging underpass use as it does in creating CZs. Suriunary My three study sites were not replicates due of the inherent variability of landscapes, but a qualitative assessment of the results indicated that CZs are related to variables from both the landscape and the local scale. Significant variables included features form both the habitat and roadway, and were measured in the field as well as from remotely acquired data. Field measurements of continuous local-scale features were, however, unlikely to yield significant results. At the large scale, the most important features type composition, slope, an.d slope complexity of, landscape surrounding the highway. At the local scale, the most important features were the location of and distance to roadside barriers, the location of drainages, and the distance froin the road to the forest edge. 128

PAGE 144

CHAPTER6 APPLYING THE RESEARCH RESULTS TO REDUCE WILDLIFEIHIGHW A Y CONFLICTS Introduction In this Chapter I discuss the relevance of the information generated by my study, as well as how to apply it. I focus Oii informing the practice of reducing high/wildlife conflicts on a project-specific basis. My research produced the following four primary fmdings: Midand large-sized mammals do not cross highways at random. Both habitat and highway design influence where these species cross the highway.' Because landscapes and highway projects are unique, there is no single foni:mla for identifying where animals cross highways. Although not formulaic, identifying crossing zones is an accessible process that combines data and knowledge about the system in question. My discussion assumes that project planners are considering up-grades to existing highways. My specific examples apply to mountainous ecosystems and mid-to large-sized mammal speCies associated with cover. However, most of the principles I discuss could'be applied to other species, in other habitat types. I begin by offering a strategy for identifying wildlife/highway conflict locations. I then provide an overview of mitigation approaches and. make recommendations about how to incol]Jorate them into highway planning. I conclude by briefly discussing 129

PAGE 145

the need to integrate highway and conservation planning in order to maximize the success of both fields in reducing wildlife/highway conflicts in the long term. Are the Results of this Study Useftll? My results indicate that midand large sized mammals do not cross highways at random, either at the local or the landscape scale. However, differences between the low and the high permeability landscapes, as well as between crossing zones and random locations, were inconsistent across the three study sites. These results were not surprising. As I discussed briefly in Chapter 5, my three study sites differed from each. other in multiple ways, including the species present, snow depths, traffic volume, highway design, and roadway footprint. However, these inconsistencies do not imply that an assessment of the surrounding habitat and highway.characteristics are useless for informing the process of reducing wildlife/highway conflicts, as I will discuss below. Given the large scale and unique nature of landscapes; opportunities to utilize strict experimental designs, in which the systems under comparison differ by only a single component, in order to draw cause/effect conclusions, are rare (Turner et al. 2001). Because ofthis basic problem, people who need information about landscape-scale process must often rely on thoughtful, qualitative assessments of data about systems of interest, in place experimental assessments. Additionally, my results serve to illustrate the importance of considering the unique variation offered by systems in order to properly understand how they function. There is an increasing realization among envirorunental scientists that natural systems are complex. and non-linear. Therefore, making generalization about their function is difficult. Rather than try to formulate a simple rule about how a natural 130

PAGE 146

process works, it is usually more accurate to answer, "it depends" (Soule and Orians, 2001). This is not to say that natural systems are chaotic, making management decisions based on their functions impossible. The variables upon which a system's function depends are accessible to professionals familiar with the system. If these professionals combine data about the system with common sense born from their experience, good management decisions can be made (Soule and Orians, 200 I). Strategies for Identifying Wildlife/Highway Conflict Locations The frrst step to reducing wildlife/highway conflicts within a specific project site is to identify the general areas where animals are most likely to come in contact with a highway (Figure 6.1). I refer to these areas as conflict zones. Conflict zones must be defmed relative to adjacent landscapes and may encompass part or all of a project site. Both project wide and site-specific mitigation measures to reduce negative wildlife/highway interactions will yield the greatest cost and biological benefits if they are implemented within confljct zones. After identifying any conflict zone that might lie within their project area, highway planners should identify the most likely crossing zones within the conflict zones. Crossing zones are the locations where animals are most likely to cross the road. They should be the focus of any site-specific mitigation measures deemed appropriate. A Strategy for Identifying Conflict Zones Conflict zones are relatively long stretches of highway(> 2000 m) that have a high probability of being crossed by wild animals. As detailed in Chapter 5, the results of my study indicate that landscapes surrounding such segments are different from the landscapes around highway segments with few crossings. At my three study sites, habitat suitability, steepness of the terrain, and the presence of linear 131

PAGE 147

guideways in the landscape all played a role in determining which sub-areas animals crossed more often. However, the importance of each factor varied because the three variables interacted among themselves and their final effect on crossing behavior was modulated by the design of the highways they surrounded. The influence of landscape scale habitat characteristics on locations where animals crossed the highway is an excellent example of the "it depends" problem. At VP and VPS, habitat suitability for mule deer, elk, and coyotes was roughly comparable, and therefore could exert little influence on where animals preferred to be. The average slope did vary between the two sub-areas of each site, and at both sites animals crossed the roadway more often in the sub-area where the surrounding habitat was less steep. This tendency was further reinforced at VPS by high quality. foraging opportunities for coyotes associated with C:MR. At TCP, the habitat in the south was highly suitable for mule deer, the most abundant species at that site, where as much of the habitat surrounding the highway in the north end was less suitable. Although the terrain was steeper in the south end, habitat suitability overrode any preference for gentler slopes. Deer were more likely to use south end, and consequently to interact with the highway there. 132

PAGE 148

w w i ofhlterest i ------1) Identify Conflict Locations Identify Loeatlons By: Combining data with local familiarity Considering both the loeal and the landscape scale Using habitot suitability as the primary indicator f i -Habitat fragmentation : -Animal mortality i Human safety -------------1--------------Mnlmlze Cost and Biological Effectiveness By: Considering mitigation constraints up-front Using standard as wei!BS specialized design components to mitigate Using design components to achieve multiple goals ... .. 2) Identify Best Mitigation t _. 3) Integrate Mitigation Planning into Highway Planning Strategies Mitigation Choices Indude: Avoiding conflict locations Managing surrounding landscape mitigation inaease at-grade crossing opportunities by decreasing barriers -include sttuctures !hot act as over-or underpasses Examining how other variables influence -use b.uriers to funnel animals to safe crossing locations habiiat preference -locate and map important features detennine abundance BDd variation of features -place greatest reliBDce on rare, variable, or preferred features Figure 6.1 A framework for mitigating wildlife/highway conflicts along existing highways.

PAGE 149

The orientation of the predominant linear topographic features in each sub-area further reinforced the patterns created by the interactions of habitat quality and slope. The effects of the linear features were in tum influenced by the highway design. At the three sub-areas with lower densities of crossing TRs {TCP north, VP west, VPS Not C:MR), the predominate ridgelines and drainages ran parallel to the road and did not act to bring animals to the roadside. The four major drainages that did bisect the highway on the west side of VP were spanned by large bridges that provided exceptional opportunities for animals to cross under the highway, further reducing the likelihood of at-grade crossing in this sub-area. Conversely, at the south end ofTCP, the major drainages that guided animals to the highway were spanned by box culverts which deer were reluctant to enter. Consequently the number of at-grade crossings in this area was high. Although the interactions that supported the observed large-scale patterns of crossing behavior were not straight forward, they are accessible to someone familiar with the resources available in the landscapes in quest,ion as well as the habitat preferences and behavior of the species under consideration. The results suggest that a common sense analysis of landscapes, which considers composition and complexity ofthat landscape in relationship to the needs of a given species, can identify the broad zones where that species is most likely to approach a highway. This type of analysis must be data-based rather than opinion-based, but the extent of local familiarity that the analyst brings to the problem is also be important. The information discussed above suggests the following strategy for identifying areas with a high potential to be conflict zones: Employ professionals familiar with the landscapes and species of concern. 134

PAGE 150

Consider habitat suitability as the primary indicator of a potential conflict zone. Consider how other components of landscape structure may interact with habitat suitability and either increase or decrease the level of use an area receives by a particular species. Consider how design of the existing highway affects the expression of habitat preference at the roadside. A Strategy for Identifying Crossing Zones Crossing zones are the. local-scale locations where midand large-sized mammals focus crossing activity. My results indicate that features of both the surrounding natural habitat and the existing highway may cue animals to cross in certain places, creating a crossing zone. However, although drainages, barriers, and distance to the forest edge showed the strongest positive relationships crossing zones locations, there was no single suite of variables associated with all crossing zones. Choosing the best variables to indicate which locations are most likely to be crossing zones is another example of an "it depends" problem. My research indicated that, just as with the landscape-scale local conditions and interactions between variables mediated the amount of influence a variable exerted at a particular study site. An important local condition that regulates whether a feature may be useful for identifying crossing zones is the amount of variability available for measurement in that feature. For example, crossing zones at four out of the six sub-areas were positively associated with locations that were closer then expected to the forest edge, and at TCP north it was the single most important variable. However, at VP west the design and construction of the roadway created a very consistent distance 135

PAGE 151

between the pavement and the forest edge. Consequently, there was little variability for animals to cue on, and distance to forest edge was not correlated with crossing zone locations at this sub-area. Other unique local conditions can also play key roles m determining if a feature may be useful for identifying crossing zones. As noted above, the positive association with a nearby forest edge was the strongest at TCP north. This relationship was so strong that it in tum created a strong negative association with drainages. Reasons for the strength of the relationship with forest edge include the following: 1) the cover type within 100 m of the roadside was mostly open grasslands at this site, creating a relatively narrow tongue of forest leading to the roadside; 2) the forest edge was generally a long distance from the roadside in this sub-area, magnifying its effect where it came in contact with the roadway; 3) there were few other well-defined features, such as drainages or barriers, which could also act to focus crossing activity, and those that existed were far away. from the forest edge. None of these three conditions existed at any of the other five subareas. Another example of unique local condition overriding other variables might otherwise act as cues to crossing is the presence of the Copper Mountain ski area at the foot ofVail Pass. In wintertime, the lure of food and easy travel was so strong for coyotes in the area that neither the locations of barriers nor distance to the forest edge was important to them when the crossed the road, and they showed a weak negative association with drainages. Additionally, they used all slope classes consistent with their availability, even though animals crossing VPS Not CMR showed a strong preference for shallow slopes. 136

PAGE 152

As with the landscape-scale variables, the relationships of the measured variables to crossing zones were not consistent. However, they were also accessible to someone familiar with the resources available in the landscapes in question, as well as the habitat preferences and behavior of the species under consideration. A combination of data and knowledge about the system should provide a reasonable basis for deciding which locations are most likely to be crossing zones. The information discussed above suggests the.following strategy for identifying locations with a high potential to be crossing zones: Locate and map features known to be associated with crossing zones and important to the species present, especially the location of drainages,. barriers, special habitat features, and the distance to the forest edge. Using these maps, determine the relative abundance of each feature, and how much variation it exhibits along the roadside. Place greater reliance on features that are highly attractive to resident species, especially if those features are rare, and to features that are relatively variable. Approaches for Reducing Wildlife/Highway Conflicts After identifying the areas within a highway project that are most likely to be conflict locations, the next step for reducing project specific impacts to midand large-sizedmammals is to choose a mitigation approach (Figure 6.'1). Because CZ locations are a function ofboth landscapeand local-scale features, strategies for reducing wildlife/highway conflicts must consider information from both scales. Below, I discuss three primary approaches; the first two approaches consider 137

PAGE 153

landscape-scale cues while the third one focuses on local-scale cues. These threee primary approaches are: Place highways in landscapes that are least likely to create conflict zones. Managing .the surrounding landscape to reduce conflicts. Utilizing highway designs that guide animals to or away from specific crossing zones. Identify and A void High Conflict Locations Some landscapes are more likely to have denser wildlife populations and/or to facilitate movement of animals to the roadside. Using the strategy described in the previous section, planners can identify those areas and avoid placing highways in them. In practice, a highway alignment must meet a variety of criteria, not only the reduction of wildlife/highway conflicts. Therefore, a least-cost analysis to determine the best location when all criteria are considered could be used to facilitate the process. Least-cost analysis tools, which can incorporate these kinds of considerations, have been developed for highway siting (e.g., Innes and Pugh, 1996; Jha, 2000) The "identify and avoid" approach is, of course, most effective when implemented prior to construction as part of the process for choosing the best alignment for a new highway. Because few completely new highways are likely to be built in the United States, the practicality of avoiding high conflict areas altogether is limited. However, upgrades and re-alignments of existing roadways provide opportunities to implement this approach on a limited basis. Upgrades may have fewer constraints, as compared with a new alignments, especially in non-urbanized areas where wildlife populations tend to be larger. Therefore the opportunity to 138

PAGE 154

(re)design with wildlife in mind might also be comparatively greater in these situations. Managing the Surrounding Landscape to Reduce Conflicts Wild animals cross roads to escape unfavorable conditions, or to access resources. Depending on the nature of the motivator, it may be possible to remove the negative stimuli or to provide the resource on both sides of the roadway, reducing the need to cross. For example, easements or agreements to protect habitat areas from disturbance and development may encourage wildlife to "stay put." Discrete resources, such as water sources or saltlicks, are identifiable and may be easy to provide on both side of a roadway. Additionally, animal populations that predominantly use one side of a highway may be encouraged to remain there by managing that landscape to improve overall quality of resources such as forage, cover, and security on the preferred side. Depending on the species of interest, land management strategies might include plantings, clearing and/or burning to promote the growth of certain or protecting habitat from disturbance to maintain .older aged stands. Certain resources, however, exist as an integral part of the landscape and it is not possible to provide them artificially. For example, there is no way to duplicate the qualities of a low elevation wintering area with predominately southern exposures at a location comprised of high elevations with predominantly northern exposures. Resource induced movements that are unlikely to be reduced through management of the surrounding habitat include finding mating opportunities, accessing traditional birthing grounds, and seasonal migrations to swnmering and wintering habitat. Wildlife must be able to cross highways in order to access these types of 139

PAGE 155

resources and may be both behaviorally and genetically predisposed to move to these resources at certain times of the year. In addition to the problems presented by unique resources that cannot be recreated through management practices, it may be difficult to induce animals that are crossing a highway routinely as part of their daily movements to "stay put." This may be the case if the resources on both side of the highway are comparable, and animals simply cross because their daily home range includes both areas. Under these it may be difficult to argue that the ability to cross the highway is a concern for the population in question. However, if A VCs present a human safety hazard in that location, or the animal population of concern is threatened by mortality, mitigation may still be warranted. Design Based Awroaches For Reducing Conflicts As detailed above, my results indicate that habitat features and the design of the highway itself influence the locations of crossing zones. In the case of mid-and large-sized species that are associated with cover in the Southern Rocky Mountains, these local-scale features include the location of roadside barriers, the location ofhighway structures (e.g., bridges, culverts) that can act as underpasses, the orientation of major linear features of the landscape, and the location of the forest edge. Special attention also need to paid to the presence and specific location of unique and attractive resources. By considering of the location of features listed above, relative to the roadway, highway designers can intentionally facilitate or discourage crossing at particular locations alongthe roadside. Careful placement of roadway structures relative to habitat features will allow animals to cross highways safely in the locations where 140

PAGE 156

they naturally approach the roadside, reducing wildlife/highway conflicts. Components of highway projects that facilitate crossing may be structures added specifically for mitigation purposes. In other cases, adjustments to design features that that would be part of a project for other reasons may be all that is necessary to meet mitigation needs. For example, over-sizing a box culvert required for drainage may create a suitable animal underpass. Extending roadside barriers placed to meet safety requirements where the drainage bisects the roadway may then help guide animals to that underpass. Design-based approaches for mitigating wildlife/highway conflicts include: In locations with low traffic volumes, increasing at-grade crossing opportunities by minimizing the barrier effect of the highway itself. Including structural components in the roadway to accommodate animal movements above or below grade. Decreasing the attractiveness of at-grading crossing away from mitigated locations to further encourage animals to take advantage of safe crossing opportunities. For low-volume highways where A VCs do not present a significant safety hazard to animals or to the occupants of vehicles, the most costand biologically effective strategy foi reducing wildlife/highway conflicts is simply to encourage animals to cross freely at grade. Limiting the use of steep cuts and fills, Jersey barrier, guardrails, and retaining walls should minimize the barrier effect of the roadway. In locations that are identified as potential crossing zones and where these types of barriers are required to meet safety standards, barrier ends should be located where there is a good line-of-sight to give motorists adequate time to avoid animals that enter the roadway at these locations. 141

PAGE 157

Strategies that discourage at-grade crossings and act to guide animals to locations with above-or below-grade crossing opportunities may be the preferred approach under a variety of circumstances. These include high-volume roads, high A VC areas, and areas where highways conflict with migration routes or the movements of threatened or endangered species. Structural components that increase permeability of the roadway include underpasses and overpasses built especially to facilitate highway crossing by wildlife as well as other structures that can serve the same purpose. Generously sized culverts, bridges with adequate headroom, and alignments that include either elevated sections of roadway or contain the roadway in a tunnel can all allow animals to safely cross a highway. Barriers that can help to guide animals to crossing locations include guardrail, Jersey barrier, retaining walls, cliffs, steep cuts and fills, and animalprooffencing. Crossing structures are unlikely to be used if placed in a location where the surrounding habitat does not naturally bring animals to the roadside. Conversely, barriers should not be placed within exist crossing zones without acc.ompanying crossing structures. With the exception of carefully maintained mesh. fence, that is buried at the toe and at least two meters in height (CleveQger et al., 200 I), barriers will not dissuade crossing if placed in locations where landscape structure combines with resource distribution to encourage crossing behavior. Breaks in barriers placed in a crossing zone may become a focal crossing location, creating a serious A VC hazard. In addition to locating crossing structures appropriately, a crucial consideration when implementing design-based mitigation approaches is to identify the location and length of conflict zones and crossing zones relative to the highway project area. Conflict zones may extend for many kilometers along a roadway and crossing zones may also be many hundreds of meters in length. Therefore a 142

PAGE 158

relatively small project area may only partially overlap with or be completely contained within either type of area. If the barriers used to guide animals to crossing locations within the project area simply end at the project's boundaries, they may funnel animals on to the roadway at these locations, potentially creating a more intense source of conflict than previously existed. In such cases, it may be necessary to extend the mitigation project beyond the boundaries of the highway project for design-based mitigation to be successful. Integrating Mitigation Planning into Highway Planning The final step for reducing the wildlife/highway conflicts associated with a particular highway project is to incorporate the chosen mitigation strategy into project planning and design (Figure 6.1) Successfully implementing mitigation projects that reduce highways/wildlife conflicts requires that wildlife considerations be integrated into the initial stages of project planning. Incorporating wildlife consideration into project pla.'llling up-front maximizes the opportunity to fmd engineering solutions that combine the project's purpose and need with conservation considerations. For example, even within the purpose and need of a project there is usually flexibility in choosing alignments and curve geometry. Both of these design components have a strong influence on size and placement of cuts, fills, and retaining walls, as well as the need for and placement of drainage structures. Choosing an alignment which will naturally include drainage structures that can be designed to act as underpasses, and barriers that can act to funnel animals towards such crossing opportunities, may be more cost effective then adding these structural components to plans completed without consideration for wildlife mitigation. 143

PAGE 159

Because the best approach for many mitigation situations may be to incorporate wildlife-oriented design considerations throughout the project area, it is inefficient to "add on" wildlife considerations after the general design of the project is already completed. Adding wildlife mitigation as an afterthought to overall project design is likely to result in inadequate design-based mitigation and lead to cost overruns incurred as a result of delays that occur due to the re-design process. Even if the fmal "wildlife friendly" design deviates only minimally from the plans created with no wildlife considerations, the evaluation process may cause significant delays, and consequent added cost. Especially when mitigation can be accomplished simply by adjusting design features that that would be part of a project for other reasons, considering mitigation requirements up front in the planning process is the most efficient approach. In summary, important considerations to maximize both the cost-effectiveness and the biological effectiveness of mitigation projects include: Considering the reduction of wildlife/highway conflicts as baseline design constraint. Taking advantage of mitigation opportunities that are part of a standard design, e.g;, bridging rather the filling and culverting drainages. Engineering solutions that do double duty, e.g., a steep cut slope that also serves as a barrier to guide animals to a safe crossing location. Improving the Practice of Reducing Wildlife/Highway Conflicts The practice of reducing wildlife/highway conflicts can be improved by identifying conflict locations at both a landscape and a local scale, considering 144

PAGE 160

conflict locations when choosing the best strategies and locations for mitigation, and integrating this process into highway project. planning up front. However, although integrating conservation considerations into highway planning is essential for reducing wildlife/highway conflicts, it is not sufficient to achieve this goal. Three other conditions will also greatly improve this practice. Highway planners and conservation planners must build a strong partnership, all entities which affect land-use must cooperate with this new partnership, and more data about wildlife/highway interactions inust be generated to inform the decisions that these new partnerships must make. To reduce wildlife/highway conflicts in the long term, highway planning must be carried out in concert with conservation planning and vise-versa. Conservation plans that address entire regions and even continents stand the best chance of successfully preserving the greatest range of species (Noss, 1992; Soule and Terborgh, 1999). Because highways already span nearly every region ofNorth America, it is essential for these types of conservation plans to consider the impacts of current highways, and upgrades that may be planned for them in the future. Additionally, because of their extensive nature, the impacts ofhighways are also an issue for many local conservation problems.lfhighway and conservation planners coordinate at both scales, highway planners can strive to reduce or even remove roadway impacts in critical conservation areas, and conservation planners can focus protection efforts in the areas least likely to suffer from highway effects. Although highway and conservation planners have historically ignored each other, the stage has been set for this type of cooperation. As reviewed in Chapter 1, both highway specific legislation and a paradigm shift in natural resource management are encouraging this new partnership. 145

PAGE 161

Because of the landscape-scale issues associated with mitigating wildlife/highway conflicts, it is imperative that highway planners coordinate with the professionals who plan for and manage the area that surrounds the highway. When highways divide areas that are primarily managed for conservation purposes, such as national forests or state wildlife management lands, highway professionals may deal directly with conservation planners. However, a wide variety of professionals, most of whom do not identify themselves as conservation planners per se, influence and regulate patterns of land-use around highways. Because land use patterns strongly influence how wildlife populations use an area, highway planners must be prepared to work collaboratively with various jurisdictions to design successful mitigation projects. Decisions that affect land use are made at all levels of government. Examples include: federal agencies designating public land as developed versus undeveloped recreation areas; states or counties acquiring and protecting open space; and county and municipal zoiling regulations. Additionally, public buy-in may be essential for mitigation projects that have the potential to impact directly private property or public lands. For example, a project could funnel large numbers ofwild animals onto private property, creating a nuisance. On public land that has high recreation value, users may object to the visual impact of proposed mitigation measures (e.g., fencing), or restrictions placed on recreational activities in order to reduce disturbance in and around the mitigation area. Therefore, both highway and conservation planners need to be prepared to create or participate in processes that allow the public to contribute to decisions related to reducing wildlife/highway conflicts. Finally, there is a need to generate additional data about wildlife/highway interactions. As discussed in Chapter 1, this type of information remains scarce. Good information forms the baseline for buy-in by all the different groups with a 146

PAGE 162

stake in the practice of reducing wildlife/highway conflicts. Informing this practice with data allows highway planners to feel confident that mitigation projects are a worthwhile investment and conservation planners to feel confident that mitigation will be biologically effective. The public and jurisdictions that control surrounding land use are also more likely to agree to match their activities to the needs of wildlife/highway conflict reduction if data indicates that there is specific reason under take a mitigation project in a particular location, and that the project will have a positive effect. Summary The features associated with conflict zone and crossing zones vary in different landscapes. Nevertheless, these variables provide an accessible source of information to improve the practice of reducing wildlife/highway conflicts. Conflict reduction should include identifying conflict locations at both a landscape and a local scale, considering conflict locations when choosing the best strategies and locations for mitigation, and integrating this process into highway project planning up front. Increased data about wildlife/highway interactions and coordination among highway and conservation planners, as well as other entities that affect land use, will improve the practice of reducing wildlife/highway conflict in the future. 147

PAGE 163

CMR CSD cz DAU DEM EAU ESR GIS GMU GPS MP NHS NLCD ROW SPSS TCP TE TR USGS U1M VP VPS APPENDIX A. ABBREVIATIONS USED IN TEXT Copper Mountain Resort Context Sensitive Design crossing zone deer analysis unit digital elevation model elk analysis unit edge-to-size ratio geographic information system game management unit global positioning system mile post National Highway System national land cover data right-of-way Statistical Package for the Social Sciences Trout Creek Pass study area transportation enhancement track record United States Geological Survey universal trans mercator Vail Pass study area Vail Pass Snow study area 148

PAGE 164

BIBLOGRAPHY Alexander, S.M., Waters, N. M., 2000. The effects ofhighway transportation corridors on wildlife: a case study of Banff National Park. Transport. Res. C-Emerg. Technol., 8, 307-320. Allen, R. E., McCullough, D. R., 1976. Deer-Car accidents in Southern Michigan. J. Wildl. Manage., 40, 317-325. Andrews, R.N. L., 1999. Managing the Environment, Managing Ourselves. Yale University Press, New Haven. Baily, T. C., Gatrell, A. C., l995.1nteractive Spatial Data. Longman Group, England. Bashore, T. L., Tzilkowski, W. M., Bellis, E. D., 1985. Analysis of deer-vehicle collision sites in Pennsylvania. J. Wildl. Manage., 49,769-774. Beier, P., 1994. Dispersal of juvenile cougars in fragmented habitat. J. Wildl. Manage. 59, 228-237. Bellis, E. D., Graves, H. B., 1971. Deer mortality on a Pennsylvania interstate highway. J. Wildl. Manage., 35, 232-341. Bennett, A. F., 1991. Roads, roadsides, and wildlife conservation: a review. In: Saunders, D. A., Hobbs, R. J. (Editors), Nature Conservation 2: The Role of Corridors. Surrey Beatty & Sons, place, pp.99-118. Besag, J., Diggle, P. J., 1977. Simple Monte Carlo tests for spatial patterns. Appl. Statistics, 26, 327-333. Born, S.M., Sonzogni, W.C., 1995. Integrated environmental management: strengthening the conceptualization. Environ. Manage., 19, 167-181. Brossard, P. F., Reed, J. M., Tracy, C. R., 1998. Ecosystem management: what is it really? Landsc. Urban Plann., 40, 9-20. Carson, R., 1951. The Sea Around Us. Oxford University Press, New York. 149

PAGE 165

Carson, R., 1955. The Edge of the Sea. Houghton Mifflin, Boston. Carson, R., 1962. Silent Spring. Houghton Mifflin, Boston. CDOT, 2002. Traffic volume website. Accessed 2002, July 1. Cottam, C., 1931. Birds and motor cars in South Dakota. Wilson Bull., 43, 313314 Carbaugh, B., Vaughan, J.P., Bellis, E. D., Graves, H. B., 1975. Distribution and activity of white-tailed deer along an interstate highway. J. Wildl. Manage., 39, 570-581. Christensen, N. L., 1996. The Report of the Ecological Society of America Committee on the Scientific Basis for Ecosystem Management. Ecol. Appl. 6, 665-691. Clevenger, A. P., Waltho, N., 2000. Factors influencing the effectiveness of wildlife underpasses in BanffNational Park, Alberta, Canada. Conserv. Biol.; 14, 47-56. Clevenger, A. P., Chruszcz, B., Gunson, K., 2001. Drainage culverts as habitat linkages and factors affecting passage by mammals. J. Appl. Ecol., 38, 1340-1349. Dale, V. H., Brown, S., Haeuber, R. A., Hobbs, N. T., Huntly, N., Naiman, R. J., Riebsame, W. E., Turner, M.G., Valone, T. J., 2000. Ecological principles and guidelines for managing the use of land. Ecol. Appl. 10, 639-670. Davis, W.H., 1934. The automobile as a destroyer of life. Science 79:504-5. Defenders of Wildlife, 2002. Habitat and Highways Campaign. http://defenders.org/habitatlhighways. Accessed 2002, February 6. Diamond, J. M., 1975. The island dilemma: lessons of modem biogeographic studies for the design of nature reserves. Biol. Conserv. 7, 129-146. 150

PAGE 166

Evink, G. L., Garret, P., Zeigler, D., Berry, J. (Editors), 1996. Proceedings ofthe Florida Department of Transportation/Federal Highway Administration Transportation Related Wildlife Mortality Seminar. Florida Department of Transportation, Tallahassee, FL Evink, G. L., Garret, P., Zeigler, D., Berry, J. (Editors), 1998. Proceedings ofthe International Conference on Wildlife Ecology and Transportation. FL-ER69-68, Florida Department of Transportation, Tallahassee, FL. Evink, G. L., Garret, P., Zeigler, D. (Editors), 1999. Proceedings of the Third International Conference on Wildlife Ecology and Transportation. FL-ER73-99, Florida Department of Transportation, Tallahassee, FL. Feldhamer, G. A., Gates, J. E., Harman, D. M., Loranger, A. J., Dixon, K. R., 1986. Effects of interstate highway fencing on white-tailed deer activity. J. Wildl. Manage., 50,497-503. FHWA, 1978.1-70 in a Mountain Environment: Vail Pass Colorado. Publication Number FHW ATST8-208. FHW A, 1997. Flexibility in Highway Design. Publication Number FHWA-PD P7-062. FHW A, 200 I. Principles of Context Sensitive Design. http://www.fhwa.dot.gov/csd/gualities.html. Accessed September 6, 2001. Finder, R. A., Roseberry, J. L., Woolf, A., 1999. Site and landscape conditions at white-tailed deer vehicle collision locations in Illinois. Landsc. Urban Plann., 44, 77-85. Findlay, C. S., Bourdages, J., 2000. Response time of wetlands biodiversity to road construction on adjacent lands. Conserv. Bioi. 14, 86-94. Fitzgerald, J.P., Meaney, C. A., Armstrong, D. M., 1994. Mammals of Colorado. Museum of Natural History and University Press of Colorado, Niwot. Forman, R. T. T., 1998. Road ecology: a solution for the giant embracing us. Landsc. Ecol. 13, iii-v. Forman, R. T. T., 2000. Estimate ofthe area affected ecologically by the road system in the United States. Conserv. Bioi. 14,31-35. 151

PAGE 167

Forman, R. T. T., 2002. Road Ecology: Science and Solutions. Island Press, Washington D.C. Foster, M. L., Humphrey, S. R., 1995. Use of highway underpasses by Florida panthers and other wildlife. Wildl. Soc. Bull., 23, 95-100. Gibeau, M. L., Clevenger, A. P., Herrero, S., Wierzchowski, J., 2002. Grizzly bear response to human development and activities in the Bow River Watershed, Alberta, Canada. Bioi. Conserv., 103,227-236. Grumbine, R. E., 1994a. What is ecosystem management? Conserv. Bioi., 8, 2738. Haas, C. D., 2000 .. Distribution, Relative Abundance, and Roadway Underpass Responses of Carnivores throughout the Puente-Chino Hills. Thesis, Biological Sciences, California State Polytechnic University, Pomona. Haeuber, R., 1996. Setting the environmental policy agenda: the case for ecosystem management. Ecol. Appl. 6, 692-693. Hansen, A. J., Spies, T. A, Swanson, F. J., Ohmann, J. L., 1991. Conserving biodiversitY in managed forests. Bioscience 41, 382-392. Hansen, A., Garman, S., Marks, B., Urban, D. L.,1993. An approach for managing vertebrate diversity across multiple-use landscapes. Ecol. Appl. 3, 481496. Hewes, L. 1., 1950. Highway engineering. In: Labatut, J., Lane, W.J. (Editors), Highways in our National Life. Amo Press, New York, pp. 326-336. Hewitt, D.G., Cain, A.,Tuovila, V., Shindle, D. B., Tewes, M. E., 1998. Impacts of an expanded highway on ocelots and bobcats in southern Texas and their preferences for highway crossings. In: Evink, G. L., Garret, P., Zeigler, D., and Berry, J., (Editors), Proceedings of the International Conference on Wildlife Ecology and Transportation. FL-ER-69-68, Florida Department of Transportation, Tallahassee, FL, pp. 10-16. Hoctor, T. S., Carr, M. H., Zwick, P. D., 2000. Identifying a linked reserve system using a regional landscape approach: the Florida ecological network. Conserv. Bioi. 14,984-1000. 152

PAGE 168

Hordequin, M., 2000. Introduction to special section: Ecological effects ofroads. Conserv. Bioi. 14, 16-17. Hubbard, M. W., Danielson, B. J., Schmitz, R. A., 2000. Factors influencing the location of deer-vehicle accidents in Iowa. J. Wildl. Manage., 64, 707-713. Hunt A., Dickens, H. J., Whelan, R. J., 1987. Movement of mammals through tunnels under railway lines. Australian Zool. 24, 89,.93. Inbar, M., Mayer, R. T., 1999. Spatio-temporal trends in armadillo diurnal activity and road-kills in central Florida. Wildt. Soc. BulL, 27, 865-872. Innes, J.D., Pugh, E. D;, 1996. A model for the environmental screening of highway projects. Can. J. Civ. Engineer., 23, 518-533. Iverson, A. L., Iverson, L. R., 1999. Spatial and temporal trends of deer harvest and deer-vehicle accidents in Ohio. Ohio J. Sci., 99, 84-94. Jha, M. K., 2000. Using a geographic information systems for automated decision making in highway cost analysis. Transport. Res. Rec. 1768,260-267. Jones, J. A., Swanson, F. J., Wemple, B. C., Snyder, K. U., 2000. Effects of roads on hydrology, geomorphology, and disturbance patches in stream networks. Conserv. Bioi. 14, 76-85. Jones, M. E., 2000. Road upgrade, road mortality and remedial measures: impacts on a population.of eastern quolls and Tasmanian devils. Wildt. Res. 27, 289-296. Kaszynski, W., 2000. The American Highway. McFarland & Company, Inc., Jefferson. Kingsland, S. E., 1991. Defining ecology as science. In: Real, L.A., Brown, J. H., (Editors), Foundations ofEcology. University of Chicago Press, Chicago. Kline, B., 1997. First Along the River: A BriefHistory of the Environmental Movement. Acada Books, San Francisco. 153

PAGE 169

Kohn, B., Frair, J., Unger, D., Anderson, E., 1997. Impacts of highway development on northwestern Wisconsin timber wolves. Annual Report, Wisconsin Department ofNatural Resources, 8 p. Kushlan, J. A., 1988. Conservation and management of the American crocodile. Environ. Manage. 12, 777-790. Lane, W. J., 1950. The early highway in America, to the coming of the railroad. In: Labatut, J., Lane, W.J. (Editors), Highways in our National Life. Amo Press, New York, pp. 66-76. Lewis, T ., 1997. Divided Highways. Penguin Putnam, New York. Leopold, A., 1933. Game Management. University of Wisconsin Press, Madison. Leopold, A., 1949. A Sand County Almanac. Oxford University Press, New York. Lidicker, W. Z., Jr., 1999. Responses of mammals to habitat edges: an overview. Landsc. Ecol.l4, 333-343. Lister, N., 1997. A systems approach to biodiversity conservation planning. Environ. Monitor. Assess. 49, 123-155 Lovallo, M. J., Anderson, E. 1996. Bobcat movements and home ranges relative to roads in Wisconsin. Wildl. Soc. Bull. 24, 71-76 Lyon, J. L., 1979. Habitat effectiveness for elk as influenced by roads and cover. J. Forestry. 77, 658-660 MacArthur, R. H., Wilson, E. 0., 1967. The Theory oflsland Biogeography. Princeton University Press, Princeton. McHarg, I. L., 1969. Design with Nature. Natural History Press, Garden City. McLellan, B. N., Shackleton, D. M., 1988. Grizzly bears and resource extraction industries: effects of roads on behavior, habitat use, and demography. J, Appl. Ecol. 25, 451-460. Martin, E. W., 2001. Cross Test, an Arc View extension writen in the Avenue programming language. CommEn Space, Seattle, WA. 154

PAGE 170

Mamalis, J., 1995. Wildlife use of the Trans-Canada Highway phase III. Technical Report, BanffNational Park. 11 p. Margules, C. R., Pressey, R. L., 2000. Systematic conservation planning. Nature 405,243-253. Mech, L.D., Fritts, S. H., Radde, G. L., Paul, W. J., 1988. Wolf distribution and road density in Minnesota. Wildl. Soc. Bull. 18, 85-87. Mitchell, I. S., 1933. Roads and road making in colonial Connecticut. In: Tercentenary Commission ofthe State of Connecticut, Yale University Press, New Haven. Mowbray, A. Q., 1969. Road to Ruin. J.B. Lippincott Compari.y, Philadelphia and New York. Myerson, D. L., 2000. Getting it right in the right-of-way: Citizen participation in context-sensitive highway design. Scenic America Action Guide. Secenic America, Wachington D.C. Neale, H. J., 1950. The highway as parkway. In: Labatut, J., Lane, W.J. (Editors), Highways in our National Life. Amo Press, New York, pp. 318-325. Noss, R. F., 1983. A regional landscape approach to maintaining biodiversity. BioScience 33:700-706. Noss, R.F., 1992. The wildlands project: land conservation strategy. Wild Earth 1, 10-25. Patton, P., 1986. Open Road. Simon and Schuster, New York. Pfister, H. P., V. Keller, H. Reck, B. Georgii., 1996. Bio-okologische Wirksamkeit von Grunbrucken uber Verkehrswege. Forschung Strassenbau und Strassenverkehrstechnik. no. 756. Bundesministerium fur Verkehr Abteilung Strassenbau, Bonn. Pickett, S. T. A., Thompson, J. N., 1978. Patch dynamics and design of nature reserves. Bioi. Conserv. 13, 27-37. 155

PAGE 171

Poiani, K. A., Baumgartner, J. V., Buttrick, S.C., Green, S. L., Hopkins, E. Ivey, G. D., Seaton, K. P., Sutter, R. D., 1998. A scale-independent, site conservation planning framework in The Nature Conservancy. Landsc. Urban Plann. 43, 143-156. Prendergast, J. R., Quinn, R. M., Lawton, J. H., 1999. The gaps between theory and practice in selecting nature reserves. Conserv. Bioi. 13,484-492. Puglisi, M. J., Lindzey, J. S., Bellis, E. D., 1974. Factors associated with highway mortality of white-tailed deer. J. Wildl. 38, 799-807. Reijnen, Foppen, R., Ter Braak, C., Thissen, J., 1995. The effects of car traffic on breeding bird populations in woodland. III. Reduction of density in relation to the proximity of main roads. J. Appl. Ecol. 32, 187-202. Reijen, R., Foppen, R., Meeuwsen, H., 1996. The effects of traffic on the density of breeding birds in Dutch agricultural grasslands. Bioi. Conserv. 75, 255260. Reilly, R E., Green, H. E., 1974. Deer mortality on a Michigan interstate highway. J. Wildl. Manage., 38, 16-19. Richter, W. L., 1995. The ABC-CLIO Companion to Transportation in America. ABC-CLIO Inc., Santa Barbara. Rittel, H., Webber, M. M., 1973. Dilemmas in a general theory of planning. Policy Sciences 4, 155-169. Robinson, J., 1971. Highways and Our Environment. McGraw-Hill, Inc., New York. Rodriguez, A., Crema, G., Delibes, M., 1996. Use of non-wildlife passages across a high speed railway by terrestrial vertebrates. J. Appl. Ecol., 33, 15271540. Rodriguez, A., Crema, G., Delibes, M., 1997. Factors affecting crossing of red foxes and wildcats through non-wildlife passages across a high-speed railway. Ecography, 20, 287-294. Romin, L. A., Bissonette, J. A., 1996a. Deer-vehicle collisions: Status of state monitoring activities and mitigation efforts. Wildl. Soc. Bull. 24, 276-283. 156

PAGE 172

Romin, L.A., Bissonette, J. A., 1996b. Temporal and spatial distribution of highway mortality of mule deer on newly constructed roads at Jordanelle Resevior, Utah. Great Basin Nat. 56, 1-12. Romney, G., 1950. The motor vehicle and the highway: Some historical implications. In: Labatut, J., Lane, W.J. (Editors), Highways in our National Life. Amo Press, New York, pp. 215-226. Roof, J., Wooding. J., 1996. Evaluation of S.R. 46 wildlife crossing. Florida Cooperative Fish and Wildlife Research Unit, U.S. Biological Service Technical Report no. 54. 36 pp. Rost, G. R., Bailey, J. A., 1979. Distribution of mule deer and elk in relation to roads. J. Wildl. Manage., 43, 634-641. Ruediger, B., 1998. Rare carnivore and highways-moving into the 21st century. In: Evink, G. L., Garret, P., Zeigler, D., andBerry, J., (Editors); Proceedings of the International Conference on Wildlife Ecology and Transportation. FL-ER -69-68, Florida Department of Transportation, Tallahassee, FL, pp. 10-16. ScoPE, 2002. WRIS web page. Accessed July 13,2002 Seely, B. E., 1987. Building the American Highway System. Temple University Press, Philadelphia. Shafer, C., 1994. Beyond park boundaries. In: Cook, E. A., van Lier, H. N. (Editors) Landscape Planning and Ecological Networks. Elsevier, Amsterdam, the Netherlands, pp. 201-223. Simberloff, D. S., Cox, J., 1987. Consequences and costs of conservation corridors. Conserv. Bioi. 1, 63-71. Singleton, P. H., Lehmkuhl, J. F., 2000.1-90 Snoqualmie Pass wildlife habitat linkage assessment. Final Report, USDA Forestry Sciences Lab Cooperative Agreement PNW -98-0513-CC. 97 pp. Slocombe, D. S., 1993. Implementing ecosystem-based management: development of theory, practice and research for planning and managing a region. Bioscience 43, 612-622. 157

PAGE 173

Soule, M. E., 1985. What is Conservation Biology? BioScience 35, 727-734. Soule, M. E., 1991. Land-use planning and wildlife maintenance-guidelines for conserving wildlife in an urban landscape. J. Am. Plann. Assoc. 57, 313323. Soule, M. E., Orians, G. H. 2001. Conservation Biology: Research Priorities for the Next Decade. Island Press, Washington D.C. Soule, M. E., Simberloff, D. S., 1986. What do genetics and ecology tell us about the design of nature reserves? Bioi. Conserv. 35, 19.,40. Spellerberg, I. F., 1998. Ecological effects of roads and traffic: a literature review. Global Ecol. Biogeog. 7, 317-333. Stamps, J. A., Buechner, M., Krichnan,V. V., 1987. The effects of edge permeability and habitat geometry on emigration from patches of habitat. Am. Nat. 129, 533-552. Stoner, T., 1930. The toll of the automobile. Science 47, 13-18. Sweanor, L. L., Logan, K. A., Homocker, M.G., 1999. Cougar dispersal patterns, metapopulation dynamics and conservation. Conserv. Biol. 14, 798-808. TerraBerns, 1998. Canada lynx in Idaho. DRAFT. A cooperative effort of the Bureau of Land Management, Idaho Department of Game and Fish, U.S. Fish and Wildlife Service, U.S. Forest Service, and Idaho Parks and Recreation. Thiel, R.P., 1985. Relationship between road densities and wolfhabitat suitability in Wisconsin. Am. Mid. Nat. 113, 404-407 Trombulak, S.C. andFrissell, C. A., 2000. Review of ecological effects ofroads on terrestrial and aquatic communities. Conserv. Bioi. 14, 18-30. U.S.C 23, 1966. Title 23, United States Code-Highways. Section 138(4)(f), preservation of parklands. U.S.C. 23, 1991. Title 23, United States Code-Highways. Sec 133(d)(2), transportation enhancements. 158

PAGE 174

U.S.C 23, 1998. Title 23, United States Code-Highways. Section 101(a)(35) U.S.C. 23, sec 133(d)(2), transportation enhancements. USDIO Fish and Wildlife Service, 2001. Short history of the refuge system. htt,p:/ /bluegoose.awr.r9 .fWs.govlhistory/over/hist-a.html. Accessed 2001, April1. USDOT, 1976. America's Highways. U.S. Government Printing Office, Washington, D.C. van der Zande, A.N, ter Keurs, W. J., van der Weijden, W.J., 1980. The impact of roads on densities of four bird species in an open field habitat evidence of a long distance effect. Bio. Conserv. 18,299-312. Waller, J., Servheen, C., 1999. Documenting grizzly bear highway crossing patterns using GPS technology. In: Evink, G. L., Garret, P., Zeigler, D., (Editors), Proceedings of the International Conference on Wildlife Ecology and Transportation. FLER73-99, Florida Department of Transportation, Tallahassee, FL, pp. 21-24. Weingroff, R. F., 1993. A peaceful campaign of progress and reform. Public Roads 57, 1-13. Western, D., 1989. Conservation without parks: wildlife in the rural landscape. In: Western, D., Pearl, M. (Editors), Conservation for the twenty-frrst century . Oxford Press, New York, pp. 158-165. Wick, J., 1995. A State Highway Project in Your Town? Preservation Trust of Vermont. Yaffee, S. L., 1999. Three faces of ecosystem management. Conserv. Bioi. 13:713-725. Yanes, M., Velasco, J. M., Suarez, F., 1995. Permeability of roads and railways to vertebratesthe importance of culverts. Bioi. Conserv., 71, 217-222. 159