EVALUATING THE BARRIER EFFECT OF A MAJOR HIGHWAY ON MOVEMENT AND GENE FLOW OF
THE NORTHERN FLYING SQUIRREL
Principle Investigators:
Joe Smith
Steven Kalinowski
Robert Long
MAY 2009 – MAY 2011
1
2
3
ABSTRACT
To be completed.
INTRODUCTION
4
This study will employ radio-telemetry and genetic techniques to describe the barrier
5
effect of a major highway corridor on a gliding arboreal rodent, the Northern flying squirrel
6
(Glaucomys sabrinus), in the Cascade Mountains in western Washington.
7
Landscape permeability and connectivity between populations of organisms is a major
8
focus of conservation biology. Habitat fragmentation, a process driven primarily by human
9
activities, is thought to be one of the top causes of species extinction. Fragmentation of large,
10
contiguous patches of habitat into numerous smaller, isolated patches may increase the risk of
11
local extinction via several mechanisms including reduced or altered connectivity between
12
populations and negative edge effects (Mills 2007). Roads and other transportation infrastructure,
13
such as railroads, are pervasive source of habitat fragmentation around the world, affecting, for
14
example, some 19% of the land area of the coterminous United States (Forman 2000).
15
Many studies have shown either direct or indirect negative effects of transportation
16
infrastructure on wildlife populations (see Spellerberg et al 1998 for review). A major direct
17
negative effect is the ability of roads to limit movement of animals across the landscape. Roads
18
may pose barriers to individual movements through direct mortality, behavioral avoidance, or by
19
acting as an impassable physical obstacle in the landscape (Forman and Alexander 1998;
20
Balkenhol and Waits 2009). At the population level, this can result in the fragmentation of large
21
habitat patches into smaller, isolated fragments between which gene flow is limited or
22
nonexistent. This so-called barrier effect has been demonstrated for species from multiple taxa
23
using a variety of techniques including field-based techniques such as mark-recapture and
24
telemetry and, more recently, molecular genetic approaches (e.g., Gerlach and Musolf 2000,
25
Keller and Largiader 2003, Epps et al 2005, Riley et al 2006, Kuehn et al 2007, Marsh et al
26
2008). Population genetic theory predicts that small, isolated populations have a greater risk of
27
losing genetic diversity and suffering the effects of inbreeding depression. Inbreeding and
28
reduced genetic diversity in small populations can lead to reduced fitness and increased
29
susceptibility to demographic and environmental stochasticity (Crnokrak & Roff 1999, Hedrick
30
and Kalinowski 2000, Reed and Frankham 2001). Therefore, a population consisting of many
31
small subpopulations restricted to isolated patches of habitat is at greater risk of extinction than a
32
single large population of equal size (Frankham et al 2002).
33
Molecular genetic techniques have shown great potential in describing fine-scale
34
population structure and are increasingly being used to detect population-level effects of habitat
35
fragmentation (Keyghobadi 2007). Genetic techniques may be the most effective tools available
36
for detecting barrier effects and evaluating applications of wildlife crossing structures. Highly
37
variable genetic markers such as microsatellites can be used to address road effects at recent
38
temporal and fine spatial scales. Genetic studies involving roads constructed as recently as 20
39
years ago have been successful in detecting barrier effects (Balkenhol and Waits 2009).
40
Molecular techniques also allow comparison between roads of differing characteristics (e.g.,
41
Gerlach and Musolf 2000, Marsh et al 2008) and between roads and other landscape features that
42
may affect genetic structure such as topography or gradients in habitat quality (e.g., Cushman et
43
al 2006). Detectable population structuring inferred from neutral markers may also be a
44
forerunner of reduced genetic diversity, as genetic variation usually responds more slowly to
45
habitat fragmentation (Keyghobadi 2007).
46
Growing concern over wildlife-vehicle collisions (WVCs) and the barrier effect has led to
47
the construction of various kinds of crossing structures designed to facilitate the safe movement
48
of animals across roads. A growing body of studies has shown that a wide variety of species use
49
these structures (e.g. Clevenger et al 2000, Clevenger and Waltho 2003, Mansergh and Scotts
50
1989, McDonald and St. Clair 2004, Ng et al 2004). Use alone, however, does not indicate that
51
these structures are successful in mitigating the fragmenting effects of roads at the population
52
level and the majority of studies lack a comparison of pre- and post-construction movement or
53
gene flow (Roedenbeck et al 2007, Corlatti 2009).
54
In terms of conservation value, the need for wildlife crossing structures should properly
55
be evaluated at the level of population demographic rates. The important conservation functions
56
of crossing structures are to reduce road mortalities and maintain migration between otherwise
57
divided subpopulations. It is desirable to know both the relative importance of the different
58
mechanisms by which roads affect the species of interest as well as the circumstances under
59
which mitigation measures can be used with success (Roedenbeck et al 2007). If crossing
60
structures are deemed necessary to either reduce mortality or facilitate movement, then the
61
success of these mitigation measures should properly be measured at the population level, that is,
62
in terms of decreased road mortality and/or increased gene flow. To rigorously test the
63
effectiveness of wildlife crossing structures, a before-after study design is needed. Although
64
studies of crossing structures have documented animals using the structures and suggested that
65
inter-patch movement and gene flow should theoretically increase (e.g. Dodds et al 2004), to my
66
knowledge, this has not yet been tested empirically.
67
This study aims to address specific questions regarding the need for and efficacy of
68
wildlife crossing structures in increasing the movement and genetic exchange of a species that
69
managers have identified as a species of conservation interest in the Washington Cascades. As
70
part of a multi-phase renovation project, a 15-mile section of a major 4-6 lane interstate highway
71
(I-90) bisecting the Cascades in West-central Washington will be retrofitted with a variety of
72
wildlife crossing structures over the next two years, thus representing a unique opportunity to
73
gather baseline data that can later be used to evaluate the success of such mitigation measures in
74
increasing the permeability of roads to wildlife. I propose to 1) test for a pre-construction barrier
75
effect of a major highway on Northern flying squirrels, 2) compare the barrier effect of the
76
highway with other linear features such as rivers and smaller roads, and 3) establish baseline data
77
that can later be used to compare pre- and post-construction genetic differentiation across I-90.
78
The goal of this study—which is part of a larger study looking at a suite of different
79
organisms including fish, amphibians, forest mesocarnivores, bears, and small mammals—is to
80
describe how a species with a novel mode of movement responds to a major highway and to
81
document the degree to which the highway has posed a barrier to genetic exchange for this
82
species. The data obtained from this study are intended to provide a baseline with which to
83
compare future data collected after crossing structures are implemented. Very few studies of
84
wildlife crossing structures have included pre-construction measurements of animal movement or
85
genetic subdivision. After construction is completed and wildlife crossing structures are in place,
86
managers can use our results to conduct a rigorous evaluation of the effectiveness of the crossing
87
structures in increasing the permeability of the highway to gene flow for flying squirrels.
88
89
LITERATURE REVIEW
90
Road Effects
91
Transportation networks, including roads, railroads and canals, are one of the most
92
ubiquitous forms of human-induced landscape alteration worldwide, and road effects probably
93
impact the majority of ecosystems in developed countries. In the United States, for example,
94
Forman and Alexander (1998) estimate that 15-20% of the coterminous land area is ecologically
95
affected by roads, and Riitters and Wickham (2003) estimate that only ~18% of the land area is
96
more than 1 km from a road. Road densities in other developed countries are significantly higher
97
than in the United States. Populations and ecosystems unaffected by roads may be the exception
98
to the rule. Transportation networks have a variety of effects on animals and populations, and
99
research on these effects has increased dramatically in the past decade. So-called road effects
100
include direct effects in the form of mortality from vehicular collisions, modification of animal
101
behavior, alteration of habitat via pollution and edge effects, loss of habitat, and fragmentation of
102
formerly continuous habitat into isolated patches. Indirect effects of roads on ecosystems are
103
numerous and include the facilitation of spread of noxious species, increased use and alteration of
104
areas by humans, and increased availability of carrion from road kills (see Spellerberg 1998,
105
Forman and Alexander 1998, and Trombulak & Frissell 2000 for review). I focus here on the
106
barrier effect and resulting population subdivision.
107
Roads are barriers to movement for many taxa, dividing formerly continuous populations
108
into smaller, partially isolated subpopulations. The relative importance of the different
109
mechanisms by which roads impact populations is critical information for managers attempting to
110
mitigate road effects (Jaeger and Fahrig 2004, Roedenbeck et al 2007). Forman and Alexander
111
(1998) suggest that although an estimated one million vertebrates are killed on roads every day in
112
the United States, the barrier effect of roads is likely a more serious threat to most populations
113
than is direct mortality associated with traffic.
114
115
Field-based Road Ecology
116
The tendency of roads to affect animal population structure through the barrier effect
117
traditionally involved either mark-recapture methods or closely tracking marked individuals to
118
characterize animals’ movement or abundance in relation to roads and detect crossings.
119
Studies that have compared various road characteristics such as width, surface, and traffic
120
amount have returned mixed results. Oxley et al (1974) conducted extensive mark-recapture
121
studies of small and medium sized mammals and found that small mammals including white-
122
footed mice (Peromyscus leucopus), eastern chipmunks (Tamias striatus), and eastern gray
123
squirrels (Sciurus carolinensis) avoided crossing all roads in the study area but that road
124
clearance (total gap in vegetation associated with the road corridor) strongly influenced rates of
125
crossing, with clearances >20 m severely restricting crossing. They concluded that a four-lane
126
divided highway may represent as great a barrier to dispersal of small forest mammals as a body
127
of water twice as wide (Oxley et al 1974). Mader (1984) measured movement of several species
128
of Carabid beetles (Carabidae) and forest-inhabiting mice (Apodemus flavicollis and
129
Clethrionomys glareolus) near four roads of varying widths in Germany. Of over 10,000 marked
130
beetles, only a handful were recaptured on the opposite side of the road of their capture and one
131
species, an interior-forest associated species, was never detected to have even penetrated the
132
roadside vegetation (Mader 1984). Although mice were never detected to have crossed 2-lane
133
paved roads of their own volition and strongly avoided crossing even a closed, unpaved road, two
134
of 14 mice translocated to the opposite side of a smaller paved road were detected to have crossed
135
back to their side of origin (Mader and Pauritch 1981, in Mader 1984). McGregor et al (2008)
136
conducted controlled translocation experiments in eastern chipmunks and white-footed mice and
137
found that each intervening road across which animals were translocated reduced the probability
138
of return by about 50%. Traffic volumes did not have a significant effect on return probability,
139
leading the authors to conclude that these species avoid the road surface itself rather than noise or
140
other emissions associated with vehicle traffic (McGregor et al 2008). Contrasting results were
141
seen in another mark-recapture translocation study of bank voles and yellow-necked mice
142
(Apodemus flavicollis) in the Czech Republic; though neither species crossed either a 2-lane or a
143
busier, divided 4-lane highway without provocation, translocated animals would return across the
144
2-lane road but not the divided 4-lane highway (Rico et al 2007). Using mark-recapture
145
techniques, Baur & Baur (1990) found that land snails (Arianta arbustorum) very rarely crossed a
146
paved road and avoided crossing even a 3 m wide unpaved road. Swihart & Slade (1984) studied
147
prairie voles (Microtus ochrogaster) and cotton rats (Sigmodon hispidus) on either side of an
148
infrequently used two-track dirt road that was later widened and paved. The road strongly
149
inhibited movement for both species and upgrading the road did not increase this effect (Swihart
150
& Slade 1984).
151
Another high-priority research area in road ecology seeks to understand the mechanisms
152
underlying the variability in species’ vulnerability to road effects. Our knowledge of how the
153
majority of species respond to roads is entirely lacking, but the limited number of species that
154
have been examined have demonstrated the breadth of variability among species—even within
155
the same taxon—in behavioral response and population vulnerability to roads. Hence, a number
156
of studies have examined differences in the barrier effect between species or taxonomic groups in
157
an attempt to improve our ability to correctly predict which species may be particularly
158
vulnerable.
159
Theory predicts that larger, wider-ranging animals with high dispersal capability may be
160
less inhibited by roads than smaller animals and those restricted to a specific habitat type
161
(stenotopic). Indeed, Hornocker and Hash (1981) found that wolverine home ranges in
162
northwestern Montana, which averaged 422 and 388 km2 for males and females, respectively,
163
were not affected by highways. Grizzly bears (Ursus arctos), however, have been shown to avoid
164
roads and preferentially select habitats with low road density (McLellan and Shackleton 1988,
165
Mace et al 1996, Ciarniello et al 2007). Compared to small mammals (Peromyscus, Tamias,
166
Sciurus), Oxley et al (1974) found the barrier effect was generally lower for medium sized
167
mammals such as raccoons (Procyon lotor), porcupines (Erethizon dorsatum), skunks (Mephitis
168
mephitis), and groundhogs (Marmota monax), some of which crossed 4-lane highways. The
169
magnitude of barrier effects may be especially high in animals with low mobility such as turtles
170
(Shepard et al 2008).
171
Differences in life history, morphology, and ecology may also explain interspecific
172
differences in the barrier effect of roads. Laurance et al (2004) found that Amazonian forest birds
173
differed in their willingness to cross a 30-40 m wide road; frugivorous and edge and gap species
174
moved relatively freely across the road compared to forest insectivores, which rarely crossed the
175
road, and solitary understory species, which avoided even overgrown sections of the road. Kerth
176
and Melber (2009), studied two species of bats in northern Bavaria, Germany, using mark-
177
recapture and telemetry and found that barbastelle bats (Barbastella barbastellus), which forage
178
in open spaces and have pointed wings adapted for pursuing insects in flight, were uninhibited by
179
a 4- to 6-lane highway and frequently crossed overhead or through underpasses whereas
180
Bechstein’s bats (Myotis bechsteinii), which glean insects from foliage and have wing
181
morphology adapted to be highly maneuverable, rarely crossed the highway and always used
182
underpasses.
183
Molecular Road Ecology
184
Recent improvements in molecular genetic approaches to population biology provide a
185
powerful tool for measuring population structure and the number of applications of molecular
186
techniques in road ecology in peer-reviewed journals has surged in the past several yeas
187
(Balkenhol and Waits 2009). Recent studies have used genetic approaches to investigate the
188
barrier effect of roads in small mammals, invertebrates, reptiles and amphibians, mesocarnivores,
189
and ungulates.
190
Gerlach and Musolf 2000 measured allele frequencies, genetic distance, and genetic
191
diversity at seven microsatellite loci in bank voles (Clethrionomy glareolus) and compared
192
unfragmented populations to populations divided by a country road, a railroad, and a highway.
193
Significant population subdivision was found for populations on either side of the highway, but
194
not the smaller country road or railroad. The authors also found an isolation-by-distance pattern
195
and concluded that the Rhine River acted as a barrier equivalent to approximately 7 km of
196
geographic separation (Gerlach and Musolf 2000). An interstate highway represents a barrier to
197
gene flow in Red-backed salamanders (Plethodon cinerius) in the Appalachian mountains, but
198
significant genetic differentiation was not detected across any of five smaller roads (Marsh et al
199
2008). Ground beetles (Carabus violaceus) in isolated patches separated by roads built as recently
200
as 31 years prior exhibited significant population differentiation as well as reduced genetic
201
diversity, in contrast to generally insignificant amounts of subdivision between populations in the
202
same patch (Keller and Largadier 2003). Kuehn et al (2006) found that differentiation between
203
populations of roe deer separated by anthropogenic barriers (a fenced motorway and local roads
204
or railroad) was significantly greater than between populations on the same side of the barriers,
205
but did not detect reduced genetic diversity as a result of isolation.
206
Riley et al (2006) combined radio-telemetry and genetic techniques to assess the barrier
207
effect of a major freeway and a busy secondary road in southern California on bobcats (Lynx
208
rufus) and coyotes (Canis latrans). Although rates of migration inferred from telemetry
209
observations and genetic assignment tests were high enough to counteract drift, significant
210
differentiation was found between populations of both species divided by the freeway, and
211
between populations of bobcats, but not coyotes, divided by the secondary road. The authors
212
hypothesize that this discrepancy may reflect a lack of reproductive success for migrants resulting
213
from territory “pile-up” at the edge of major roads creating an additional social and behavioral
214
barrier (Riley et al 2006). This study illustrates the importance of actually measuring genetic
215
differentiation as opposed to inferring it from observations of movement and demonstrates the
216
utility of combining techniques to learn about the mechanisms underlying a barrier effect.
217
Landscape genetics approaches (see Manel et al 2003, Storfer et al 2007) have also
218
provided some insight into the barrier effect of transportation infrastructure, and allow the
219
importance of anthropogenic barriers to be weighed against other landscape features such as
220
topography, land use, and natural landscape heterogeneity. Liu et al (2009) examined factors
221
influencing population structure in Yunnan snub-nosed monkeys (Rhinopithecus beiti), which are
222
restricted to forested habitat in the Tibetan plateau. Five distinct subpopulations were inferred
223
from microsatellite analysis and habitat discontinuities, including anthropogenic habitat gaps such
224
as cultivated land and roads, explained a greater proportion of the genetic variation than
225
geographic distance (Liu et al 2009). Though the effect of roads on population genetic structure in
226
black bears (Ursus americanus) in northern Idaho was equivocal (Cushman et al 2006), roads and
227
associated development were a significant factor in isolation of subpopulations of grizzly bears
228
(Ursus arctos) in the northern Rocky mountains (Proctor et al 2005). Because of the large
229
geographical scale of these studies, it may be difficult to separate the effects of roads per se from
230
general anthropogenic land use and disturbance, which is often correlated with road presence.
231
…
232
In comparison, populations of white-footed mice in Indiana separated by 500 to 2000 m
233
of agricultural matrix showed no higher population differentiation than populations separated by
234
similar distances in continuous habitats (Mossman and Waser 2001).
235
Flying Squirrels as a model species
236
The Northern flying squirrel, Glaucomys sabrinus, is found in forested landscapes across
237
much of North America. Hall (1991) recognized 24 subspecies of G. sabrinus over its range.
238
Subspecies in the western United States and Canada are associated with coniferous forest, and
239
some have suggested that they are an indicator or possibly a keystone species for late-seral (“old-
240
growth”) forest communities (Carey 2000, but see Smith et al 2005). This notion is supported by
241
several aspects of flying squirrels’ecology. First, their diet consists largely of hypogeous
242
mycorrhizal fungi, which are important for nutrient uptake and growth of conifers and which
243
flying squirrels likely help to disperse (Carey 1995, Carey 2000, Lehmkuhl et al 2004). Second,
244
flying squirrels are an important prey species for old-growth obligate carnivores such as fishers,
245
martens, and Northern Spotted Owls (Carey 2000). Finally, several studies have shown densities
246
of flying squirrels to be significantly higher in late-seral forests than in younger forests (Carey
247
1995, Carey et al 1999, Waters and Zabel 1995).
248
Dispersal ability of northern flying squirrels is poorly documented. Dispersal and
249
movement habits with regard to habitat configuration have been studied, however, in a very
250
similar gliding mammal, the Siberian flying squirrel (Pteromys volans), in Eurasia. Gliding
251
ability of the two species is very similar. Vernes (2001) measured gliding ability of northern
252
flying squirrels in a 50-70 year old 2nd growth mixed hardwood-coniferous forest stand in New
253
Brunswick. Glides for males and females combined (n = 100) had a mean of 16.4 m, and ranged
254
from 3.2 to 45 m. This is similar to the gliding performance of P. volans, found in forests of
255
northern Eurasia and Japan (Asari et al 2007).
256
In Scandinavia, dispersal ability of P. volans in fragmented habitat was examined by
257
Selonen and Hanski (2003, 2004). Dispersing juveniles tended to disperse through preferred
258
habitat (breeding habitat) and 74.1% ± 25.7% of movements during dispersal were in preferred
259
habitat (2004). Matrix composed of good movement habitat and poor movement habitat were also
260
crossed, but open areas that could not be crossed in a single glide were almost always avoided
261
(2004). During 90-minute tracking periods, adults traveled short distances and spent little time in
262
matrix habitat, although males tended to cross more edges, travel longer distances in matrix
263
habitat, and spend more time in matrix habitat than females (2003). One male was observed to
264
cross a field 70 m wide in a single glide several times, and only one female crossed a gap wider
265
than 50 m. In general, dispersing juveniles were able to disperse long distances through
266
fragmented forests by staying mainly in preferred habitat and crossing matrix habitat only when it
267
was narrow (<150 m) or difficult to circumvent and gaps (nonhabitat) only if they could be
268
crossed with a single glide or by moving through scattered trees or bushes (2004).
269
270
APPROACH
271
Question 1) To what degree does I-90 currently represent a barrier to movement for individual
272
squirrels?
273
Radio Telemetry Approach
274
To examine the effect of I-90 on the movement of individual squirrels, I will use radio-
275
telemetry to track nightly movements and describe the home ranges of squirrels living near the
276
highway. Telemetry will enable us to test the null hypothesis that squirrels move randomly with
277
respect to the highway against the alternative hypothesis that squirrels avoid crossing the
278
highway. If crossings are observed, telemetry will also provide information about how squirrels
279
are moving across the highway (i.e., in culverts, gliding across, running across) and will enable us
280
to roughly assess the frequency of such events. If consistent patterns emerge, a rough model
281
could be developed to predict points along the highway corridor where crossings are likely
282
occurring. This sort of model would be especially helpful for managers if genetic analysis
283
indicates a need for mitigation measures for this species.
284
To the extent possible using a homing technique with ground-based radio-telemetry, we
285
will record locations of tracked squirrels at one-hour intervals during each tracking session. The
286
resulting data will consist of several “bursts” of 1 hr-spaced locations clustered temporally by
287
individual tracking sessions but spread evenly over the 2-3 month monitoring period. A straight
288
line drawn between a sequential pair of locations during a tracking session can be thought of as a
289
movement vector. If we are able to record, for example, 4 locations per squirrel per night on 10
290
nights over the monitoring period, we will have 30 separate movement vectors per squirrel. A test
291
for crossing avoidance will compare the null hypothesis, H0: squirrel movements are random with
292
respect to the highway, to the alternative hypothesis, HA: squirrels avoid crossing the highway. If
293
squirrels tend to avoid crossing the highway, then we would expect that the number of vectors
294
crossing the highway would be smaller than if the same vectors were arranged randomly within a
295
“null home range,” a space with a center at the squirrel’s day den and a radius of the maximum
296
distance the squirrel was observed from the day den.
297
Data obtained from telemetry will also function as a corroborator for the genetic analysis.
298
Riley et al (2006) combined these techniques in their study of bobcats and coyotes in a southern-
299
Californian landscape fragmented by a major freeway. While telemetry data suggested a
300
moderate rate of movement across the freeway, genetic analysis revealed a substantial degree of
301
population subdivision consistent with much lower crossing rates. The researchers hypothesized
302
animals crossing the freeway were not breeding—and therefore not contributing to gene flow—
303
because of territory ‘pile-up’ at the edges of the freeway creating an additional and associated
304
social and behavioral barrier to gene flow (Riley et al 2006). Thus, combining techniques can
305
uncover biologically significant phenomena that a single technique might fail to reveal.
306
307
Question 2) Has I-90 represented a barrier to gene flow for Northern flying squirrels in the
308
Cascades and, if so, to what degree are populations on either side of the highway genetically
309
differentiated?
310
Genetic Approach
311
We will address the issue of gene flow by estimating population subdivision in the study
312
area on either side of I-90. Models of genetic drift predict that allele frequencies among isolated
313
subpopulations will, over time, diverge due to drift acting randomly and independently on each
314
finite subpopulation. Rate of genetic differentiation among subpopulations is determined by the
315
rate of gene flow or migration—the two terms used interchangeably and usually denoted by m,
316
generation time for the organism, and the effective population size (Ne) (Wright 1940).
317
Several parameters are used to describe genetic differentiation between subpopulations
318
including FST (Wright 1951) GST (Nei 1973), RST (Slatkin 1995), and θ (Weir and Cockerham
319
1984). These parameters are all somewhat analogous in what they describe, but each makes a
320
slightly different set of assumptions regarding the genetic and demographic processes occurring
321
in the populations of interest and the sampling method used (see Holsinger and Weir 2009 for a
322
review). In reality, it is often difficult to ascertain which statistic is most appropriate for a given
323
organism or study system because the aforementioned processes are unknown. Simulation studies
324
have generally shown that the various estimators perform similarly for most data sets (e.g.,
325
Kalinowski 2002) so for the sake of brevity I will only make reference to FST in the following
326
sections although in the analysis of my data I will calculate values for each of these related
327
statistics for comparison.
328
329
Briefly, the parameter FST is defined as the proportion of total heterozygosity that is due
to differences in allele frequencies among subpopulations and is calculated as follows:
FST 
330
HT  H S
HT
331
where HT is the expected heterozygosity in the total population assuming no structure, HS is the
332
mean of expected heterozygositiesover all subpopulations. For a single locus,
HT  1  2 pii2
333
334
n
 k

2
H

c
and S  j 1   p ij 
 i1 

j 1
335
where there are k alleles (i=1, …, k), n subpopulations (j=1, …, n), cj is the relative proportion of
336
the jth subpopulation, and p
ij denotes the frequency of the ith allele in the jth subpopulation. The
337
second term in the equation for HS is the expected heterozygosity in the jth subpopulation.
338
339
To calculate FST using multiple loci, HS and HT can be determined separately for each
locus and averaged together to produce HS and HT . Then the multilocus version of FST is
340


FST 
HT  H S
HT
341
FST ranges from 0 to 1—0 indicating a perfectly panmictic population and 1 indicating
342
fixation for different alleles atevery locus (no shared alleles), which could hypothetically occur
343
between completely isolated subpopulations of finite size with no gene flow. Sewall Wright, the
344
pioneer of so-called F-statistics, demonstrated that very little gene flow is needed to counteract
345
the tendency of genetic drift to differentiate subpopulations regardless of the size of the
346
population (1969). Genetic markers should therefore be highly variable, mutate rapidly, and
347
should be free of selective forces to detect fine-scale or recent differentiation between large
348
populations. Microsatellites are therefore ideal markers for detecting population differentiation
349
from even fairly recent disruptions in gene flow and have generally replaced other markers for
350
this purpose (Halliburton 2004).
351
Our sampling scheme for genetic data is a paired control-impact design. We will obtain
352
genetic samples from squirrels at three or more paired trap sites, each pair consisting of one trap
353
site on either side of the highway. This paired trapping design will allow us to compare genotype
354
samples among trap sites on the same side as well as on opposite sides of the highway.
355
Comparison among sites on the same side will function as a control for geographic distance.
356
Analysis of multilocus microsatellite genotypes will proceed as follows. Genotypes will
357
be analyzed for deviations from Hardy-Weinburg equilibrium, linkage equilibrium, and presence
358
of null alleles using methods described in Raymond and Rousset (1995). Genetic differentiation,
359
FST, between all pairs of sites will be estimated using Weir and Cockerham’s (1984) statistic θ.
360
Two regression analyses will be conducted to test for spatial relationships of genetic distance. To
361
test for a simple isolation-by-distance effect, genetic distance FST or, alternatively, FST/(1-FST) as
362
recommended by Rousset (1997) will be plotted against the log of geographic distance between
363
sites. The second regression will include in the independent variable a binary term for the
364
presence or absence of the highway between pairs of sites such that the slope of the regression is
365
FST/(1-FST) = β1log(geographic distance) + β2(presence of highway) + c
366
where the presence of the highway between two sites is denoted by a 1 and its absence denoted by
367
a 0 in the second term of the right-hand side of the equation. The term c is a constant (y-
368
intercept). If the highway does act as a barrier, the coefficient β2 will be positive and significantly
369
different than 0.
370
MEASUREMENT METHODS
371
Trapping methods
372
We will trap flying squirrels at six to eight locations representing three or four pairs of
373
sites along the east-west highway corridor. Within the selected sites, trapping will be
374
opportunistic, the goal being merely to gather a sufficient number of genetic samples and to
375
distribute collars among individuals with home ranges directly adjacent to the highway. Suitable
376
trap sites will be selected in places with favorable conditions for Northern flying squirrels within
377
2 km of the highway along the 15-mile project area. We will first establish a loose trap grid or
378
trap line (depending on dimensions of the site) with an approximately 40 m spacing between
379
points that generally covers the favorable habitat at each site. We will place a maximum of thirty-
380
six trap points at each site. Pairs of Tomahawk 201 live traps (Tomahawk, WI, USA) will be set
381
at each grid point—one trap will be placed on the ground and the other fastened to the bole of a
382
tree at approximately chest-height (1.5 m). Traps will be covered with a close-fitting waxed
383
cardboard sleeve and bark or other forest floor debris to conceal the traps and provide entrapped
384
animals some thermal cover and protection from predators and precipitation. A handful of
385
polyester stuffing in each trap will provide nesting material. Traps will be baited with a mixture
386
of rolled oats, peanut butter, and molasses.
387
We will set traps shortly before dusk and check open traps the following morning. For
388
each flying squirrel captured we will record sex, age class (young of the year, young adult, or
389
mature adult), breeding status, weight, geographic coordinates of the trap (UTM, NAD83),
390
processing time, a unique identifier for the envelope containing the genetic sample, PIT tag
391
number, and, if applicable, collar frequency.
392
DNA samples will come from epithelial cells from cheek swabs. Two samples will be
393
collected from each individual—one using a synthetic swab and one using a cotton swab. Swabs
394
will be inserted into the mouth and scraped along the inner cheek with a twisting motion. Both
395
swabs will be placed in the same sealed sample envelope and stored in an airtight container with
396
desiccant as soon as possible following trapping.
397
We will use 4-gram PD-2C VHF neck collar transmitters manufactured by Holohil
398
Systems Ltd. (Carp, Ontario, Canada). Attachment will be around the neck following standard
399
procedures provided by the manufacturer. Antenna cables will be trimmed to avoid interference
400
with the squirrels’ tails.
401
Telemetry methods
402
I will restrict capture effort to within a typical home range length of the highway so
403
transmitters are placed on squirrels who’s movement is likely to be affected by the highway
404
corridor. I will deploy ten transmitters each year of the study for a total of twenty radio-collared
405
individuals. I will attempt to distribute transmitters in such a way that 2-3 squirrels with
406
overlapping or adjacent home ranges can be monitored simultaneously at each site.
407
Trapping for transmitter deployment will take place during late spring 2009 and 2010,
408
from late May until mid to late June. Tracking will commence as soon as a sufficient number of
409
animals are collared—probably around mid June. Each squirrel will be tracked periodically until
410
30-40 locations have been recorded including night activity locations and day-den locations.
411
Observers with handheld telemetry receivers and yagi antennas will select 2-3 squirrels on a
412
given night and attempt to locate each squirrel at one-hour intervals throughout their activity
413
cycle. The tracking schedule will emphasize obtaining an even distribution of locations over the
414
duration of the nightly activity cycle to minimize any major gaps due to rythmicity in the
415
squirrels’ movement patterns—for example, where the squirrel tends to visit a particular area at a
416
certain time of night.
417
Observers will be trained to home in on collared squirrels until they are able to either i)
418
establish visual contact with the squirrel or identify the specific tree in which the squirrel is
419
located, ii) home in to within an estimated 20 m of the squirrel, or iii) home in to within an
420
estimated 40 m of the squirrel. Location quality will be classified as 1, 2, or 3 corresponding to
421
these three situations. The latter classifications will likely be warranted where signal bounce is a
422
problem, when squirrels are high up in trees, or when observers are unable to follow a squirrel
423
into an area and must triangulate a location. Observers will record for each location the date,
424
time, transmitter frequency, location quality, UTM coordinates, bearing-to-signal (if triangulation
425
is used), and a location description and other relevant notes.
426
Microsatellite Methods
427
428
Need to consult S. Kalinowski. and N. Vu.
Statistical Methods: Radio Telemetry
429
To evaluate the degree to which the highway affects individual squirrels’ movements, I
430
will test the null hypothesis H0: squirrels move randomly with respect to the highway, against the
431
alternative hypothesis HA: squirrels avoid crossing the highway using methods modified from
432
those of Shepard et al (2008).
433
First, all location data for each squirrel will be converted into movement vectors with a
434
starting point and ending point corresponding to sequential locations (locations measured
435
approximately 1 hour apart during the same night). These vectors can then be simplified further
436
by specifying only the origin (coordinates of the starting point), angle (from true North or another
437
reference direction), and length (Euclidean straight-line distance between starting point and
438
ending point).
439
These simplified vectors will be plotted on a map and overlaid with the location of the
440
highway. The proportion of vectors crossing the highway will be used as a test statistic. A
441
randomization procedure will be used to derive a distribution of this test statistic under the null
442
hypothesis of random movement with respect to the highway. The randomization procedure will
443
use real vector lengths from the data, but will randomize angles. To simulate random movement
444
under the constraints of biological reality, starting points will also be chosen randomly, but will
445
be restricted to a “null home range” with a radius equal to the greatest distance a the squirrel was
446
observed from a day den during telemetry.
447
The probability of the observed number of highway crossings under the null hypothesis
448
will be obtained by measuring the area of the random distribution equal to or less than the test
449
statistic (Manly 1991).
450
Statistical Methods: Microsatellite Analysis
451
Significance of the genetic distance-geographic distance regression will be evaluated with
452
a simple Mantel test (Mantel 1967) and the significance of partial correlations in the multiple
453
regression will be evaluated with a partial Mantel test (Smouse et al 1986) using Pearson’s
454
product-moment correlation coefficient as the test statistic and assigning α levels using sequential
455
Bonferroni adjustments (Rice 1989). Mantel tests are used to test for significant correlation
456
between two or more matrices. The coefficients from a multiple regression are compared to those
457
generated by many random permutations of the predictive variables. The response matrix, Y,
458
contains some measure of genotypic or phenotypic “distance” between pairs of individuals or
459
populations and the predictive matrices, X1, X2…Xn contain some measure of environmental
460
distance between these same pairs. The predictive distances I am interested in are a) geographic
461
distance, Xg and b) presence of barriers between pairs of sites, Xb. The response matrix Y is
462
composed of pairwise FST values between sites. Partial Mantel’s tests (Smouse et al 1986) are
463
used to evaluate the relative contributions of more than one predictive matrix (e.g., the amount of
464
variation explained by the presence of a barrier after controlling for geographic distance). A
465
correlation coefficient, rYX, between the predictive matrix (or matrices) and response matrix is
466
obtained through calculations described in Mantel (1967) and Smouse et al (1986) and the
467
significance level is tested with a Monte Carlo randomization method by randomly permuting the
468
values in the Y matrix and calculating rYX values many times to obtain a null distribution of the
469
test statistic (Manly 1991).
470
471
REFERENCES
472
Asari Y, Yanagawa H, and Oshida T (2007) Gliding ability of the Siberian flying squirrel
473
Pteromys volans orii. Mammal Study 32: 151—154
474
Carey AB, Thysell DR, Villa LJ, et al (1996) Foundations of biodiversity in managed Douglas-fir
475
forests. In: The Role of Restoration in Ecosystem Management (Pearson DL & Klimas CV, eds).
476
Society for Ecological Restoration, Madison, pp. 68-82
477
Carey AB, Kershner J, Biswell B, et al (1999) Ecological scale and forest development:
478
squirrels, dietary fungi, and vascular plants in managed and unmanaged forests. Wildlife
479
Monographs 142: 1—71
480
Carey, AB (1995) Sciurids in Pacific Northwest managed and old-growth forests. Ecological
481
Applications 5: 648—661
482
Carey AB (2000) Ecology of Northern flying squirrels: implications for ecosystem management
483
in the Pacific Northwest, USA in Goldingay, RL and Sheibe, JS (eds.) Biology of gliding
484
mammals, Filander Verlag, Fu ̈rth, pp. 45–66
485
Carey AB (2000b) Effects of new forest management strategies on squirrel populations.
486
Ecological Applications 10: 248—257
487
Ciarniello, L. M., M. S. Boyce, D. C. Heard, and D. R. Seip. (2007) Components of grizzly bear
488
habitat selection: density, habitats, roads, and mortality risk. Journal of Wildlife Management
489
71:1446–1457.
490
Clevenger A. P., Chruszcz B. and Gunson K. (2001) Drainage culverts as habitat linkages and
491
factors affecting passage by mammals. Journal of Applied Ecology 38, 1340-1349.
492
Clevenger A. P. and Waltho N. (2000) Factors influencing the effectiveness of wildlife
493
underpasses in Banff National Park, Alberta, Canada. Conservation Biology 14, 47-56.
494
Clevenger A. P. and Waltho N. (2005) Performance indices to identify attributes of highway
495
crossing structures facilitating movement of large mammals. Biological Conservation 121, 453-
496
464.
497
Corlatti, L., Hacklander, K., Frey-Roos, F. 2009. Ability of wildlife overpasses to provide
498
connectivity and prevent genetic isolation. Conservation Biology 23: 548-556.
499
Crnokrak, P. and Roff, D. A. 1999. Inbreeding depression in the wild. Heredity 83: 260—270
500
Cushman, S. A., McKelvey, K. S., Schwartz, M. K. 2006. Gene flow in complex landscapes:
501
testing multiple hypotheses with causal modeling. The American Naturalist 168: 486-499.
502
Forman, R. T. T. 2000. Estimate of the area affected ecologically by the road system in the
503
United States. Conservation Biology 14, 31-35.
504
Forman, R. T. T. and Alexander, L. E. (1998) Roads and their major ecological effects. Annual
505
Review of Ecology and Systematics 29: 207—231
506
Frankham, R., Ballou, J. D., and Briscoe, D. A. (2002) Introduction to conservation genetics.
507
Cambridge University Press, Cambridge, UK: 617 pp.
508
Gerlach, G. and Musolf, K. 2000. Fragmentation of landscape as a cause for genetic subdivision
509
in bank voles. Conservation Biology 14: 1066-1074.
510
Hornocker, M. G., and Hash, H. S. 1981. Ecology of the wolverine in northwestern Montana.
511
Canadian Journal of Zoology 59: 1286—1301
512
Keller, I. and Largiader, C. 2003. Recent habitat fragmentation caused by major roads leads to
513
reduction of gene flow and loss of genetic variability in ground beetles. Proc. R. Soc. Lond. B.
514
270: 417–423.
515
Kuehn, R., Hindenlang, K., Holzgang, O., Senn, J., Stoeckle, B., Sperisen, C. 2007. Genetic
516
effect of transportation infrastructure on Roe deer population (Capreolus capreolus). Journal of
517
Heredity 98: 13-22.
518
Lehmkuhl, JF (2004) Epiphytic lichen diversity and biomass in low-elevation forests of the
519
eastern Washington Cascade range, USA. Forest Ecology and Management 187: 381—392
520
Lehmkuhl JF, Gould LE, Casares E, Hosford DR (2004) Truffle abundance and mycophagy by
521
northern flying squirrels in eastern Washington forests. Forest Ecology and Management 200:
522
49—65
523
Lehmkuhl JF, Kistler KD, Begley JS, et al (2006) Demography of northern flying squirrels
524
informs ecosystem management of Western interior forests. Ecological Applications 16: 584—
525
600
526
Mace, R. D., J. S. Waller, T. L. Manley, L. J. Lyon, and H. Zurring. (1996) Relationships among
527
grizzly bears, roads, and habitat in the Swan Mountains, Montana. The Journal of Applied
528
Ecology 33:1395–1404.
529
Mader, H. J. 1984. Animal habitat isolation by roads and agricultural fields. Biological
530
Conservation 29: 81—96
531
Manly, B. F. J. 1991. Randomization and Monte Carlo methods in biology. Chapman and Hall,
532
New York, USA. 281 pp.
533
Mansergh I. M. and Scotts D. J. 1989. Habitat continuity and social organisation of the mountain
534
pygmy-possum restored by tunnel. Journal of Wildlife Management 53, 701-707.
535
Mantel, N. A. (1967) The detection of disease clustering and a generalized regression approach.
536
Cancer Research 27: 209—220
537
Marsh, D. M., Page, R. B., Hanllon, T. J., et al. 2008. Effects of roads on patterns of genetic
538
differentiation in red-backed salamanders, Plethodon cinereus. Conservation Genetics 9: 603-
539
613.
540
McDonald W. and St Clair C. C. 2004. Elements that promote highway crossing structure use by
541
small mammals in Banff National Park. Journal of Applied Ecology 41, 82-93.
542
McLellan, B. N., and D. M. Shackleton. (1988) Grizzly bears and resource-extraction industries:
543
effects of roads on behavior, habitat use and demography. Journal of Applied Ecology 25:451–
544
460
545
Mills, L. S. (2007) Conservation of wildlife populations: demography, genetics, and
546
management. Blackwell Publishing, Malden, MA, USA: 407 pp.
547
Ng S. J., Dole J. W., Sauvajot R. M., Riley S. P. D. and Valone T. J. 2004. Use of highway
548
undercrossings by wildlife in southern California. Biological Conservation 115, 499-507.
549
Oxley D. J., Fenton M. B., Carmody G. R. 1974. The effects of roads on populations of small
550
mammals. J. Appl. Ecol. 11:51–59
551
Reed, D. H. and Frankham, R. (2001) How closely correlated are molecular and quantitative
552
measures of genetic diversity: a meta-analysis. Evolution 55: 1095—1103
553
Roedenbeck, I. A., Fahrig, L. Findlay, C. S., et al. 2007. The Rauischholzhausen Agenda for road
554
ecology. Ecology and Society 112: 11.
555
Selonen, V and Hanski, IK (2006) Habitat exploration and use in dispersing juvenile flying
556
squirrels. Journal of Animal Ecology 75: 1440—1449
557
Selonen, V and Hanski, IK (2004) Young flying squirrels (Pteromys volans) dispersing in
558
fragmented forests. Behavioral Ecology 15:564—571
559
Selonen, V and Hanski, IK (2003) Movements of the flying squirrel Pteromys volans in corridors
560
and in matrix habitat. Ecography 26: 641—651
561
Shepard, D. B., Kuhns, A. J., Dreslik, M. J., et al. 2008. Roads as barriers to animal movement in
562
fragmented landscapes. Animal Conservation 11: 288—296
563
Slatkin, M. 1995. A measure of population subdivision based on microsatellite allele frequencies.
564
Genetics 139: 457-462.
565
Smith, WP, Gende, SM, and Nichols, JV (2005) The northern flying squirrel as an indicator
566
species of temperate rain forest: test of an hypothesis. Ecological Applications 15: 689—700
567
Smouse, P. E., Long, J. C., and Sokal, R. R. (1986) Multiple regression and correlation extentions
568
of the Mantel test of matrix correspondence. Systematic Zoology 35: 627—632
569
Spellerberg, I. F. 1998. Ecological effects of roads and traffic: a literature review. Global Ecology
570
& Biogeography Letters 7, 317—333.
571
Swihart, R. K. and Slade N. A. 1984. Road crossing in Sigmodon hispidus and Microtus
572
ochrogaster. Journal of Mammalogy 65: 357—360
573
Trombulak, S. C. and Frissell, C.A. 2000. Review of ecological effects of roads on terrestrial and
574
aquatic communities. Conservation Biology 14(1): 18—30.
575
Vernes K (2001) Gliding performance of the northern flying squirrel (Glaucomys sabrinus) in
576
mature mixed forest of eastern Canada. Journal of Mammalogy 82: 1026—1033
577
Waters JR and Zabel CJ (1995) Northern flying squirrel densities in fir forests of northeastern
578
California. Journal of Wildlife Management 59: 858—866
579
Weir, B. S. and Cockerham, C. C.. 1984. Estimating F-statistics for the analysis of population
580
structure. Evolution 38: 1358-1370.
581
Wright, S. 1940. Breeding structure of populations in relation to speciation. American Naturalist.
582
74: 232-248.
583
Wright, S. 1951. The genetical structure of populations. Annals of Eugenics 15: 323-354.