Spatial Analysis of Road Permeability:
from Patterns to Processes
Thesis submitted in partial fulfillment of the requirements for the degree of
“DOCTOR OF PHILOSOPHY”
by
Yoav Avneon
Submitted to the Senate of Ben-Gurion University
of the Negev
18-March-2014
Beer-Sheva
Spatial Analysis of Road Permeability:
from Patterns to Processes
Thesis submitted in partial fulfillment of the requirements for the degree of
“DOCTOR OF PHILOSOPHY”
by
Yoav Avneon
Submitted to the Senate of Ben-Gurion University
of the Negev
Approved by the advisor
Approved by the Dean of the Kreitman School of Advanced Graduate Studies
18-March-2014
Beer-Sheva
This work was carried out under the supervision of
Prof. Yaron Ziv
The Life Sciences Department, Faculty of Natural science
Ben-Gurion University of the Negev
Research-Student's Affidavit when Submitting the Doctoral Thesis for Judgment
I Yoav Avneon, whose signature appears below, hereby declare that
(please mark the appropriate statements):
___ I have written this Thesis by myself, except for the help and guidance offered by my
Thesis Advisors.
___ The scientific materials included in this Thesis are products of my own research, culled
from the period during which I was a research student.
___ This Thesis incorporates research materials produced in cooperation with others,
excluding the technical help commonly received during experimental work. Therefore, I am
attaching another affidavit stating the contributions made by myself and the other participants
in this research, which has been approved by them and submitted with their approval.
Date: 18 March 2014
Student's name: Yoav Avneon
Signature: ______________
Acknowledgments
Words can only cage true meaning, yet I will try to use them the best way I can, with an attempt to
really thank my mentors, which have guided me through this whole process of my Ph.D.
Yaron, thank you for the freedom in my research and for the availability and willingness to help when
I needed. Thank you for the trust and support along the road. But most of all, thank you for sharing your
knowledge and values with me.
A huge thank to all my lab colleagues for great times, for collaborative work and for willingness to
help and free time, when much needed. I learnt so much from each and every one of you.
A big thank also to "The Ran Naor Foundation". For all of their great work in the field of road safety
and for their support in this study.
To my family, Maayan, Geffen and Chica: nothing exists without you. Thanks for the support,
encouragement and understanding. Thank you for the patience and listening ear. Thank you for being.
Words can only cage true meaning, and usually when you wish to say something, much like in
science, someone already said it better. So, I will end my acknowledgments with a quote:
"We do not inherit the earth from our ancestors, we borrow it from our children"
~ Native American Proverb
Table of Contents
Abstract .............................................................................................................................................I
Chapter 1: Introduction .................................................................................................................... 1
Aims and main study questions .................................................................................................... 6
Hypotheses and predictions .......................................................................................................... 6
Chapter 2: Description of research ................................................................................................... 9
Study area ..................................................................................................................................... 9
Data collection .............................................................................................................................. 9
Data analysis...............................................................................................................................13
Chapter 3: Results ..........................................................................................................................20
Road-kill patterns .......................................................................................................................20
Birds .......................................................................................................................................23
Mammals ................................................................................................................................31
Reptiles ...................................................................................................................................39
Passage-use rate patterns ...........................................................................................................47
Predation pressure patterns .........................................................................................................56
Chapter 4: Discussion .....................................................................................................................63
References ......................................................................................................................................74
Appendix ........................................................................................................................................84
Abstract
Roads and traffic induce landscape fragmentation and create barriers to wildlife movement. Roadinduced fragmentation has been shown to be harmful to various faunal groups, including invertebrates,
amphibians, reptiles, birds and mammals. These harmful effects impose significant ecological costs,
including habitat loss and increased within- and between-population isolation, wildlife mortality, reduced
access to resources, air, ground and water pollution, and disruption of ecological processes. Roads also
promote high rates of wildlife-vehicle collisions, one of mankind’s most evident road-related sources of
impact upon wildlife. A growing body of literature in the field of road ecology suggests that roadinduced fragmentation can be a major source of vertebrate mortality and thus potentially limit wildlife
populations. In turn, recent studies conclude that roads may pose the greatest mortality threat to wildlife.
Because traffic on roads acts more like a filter to movement rather than an absolute barrier, this effect is
commonly described as relative road permeability: the higher the movement rate across it, the greater the
permeability of the road. It is well documented that the road permeability can alter animal assemblage
composition, create patchy-populations, reduce biological diversity and increase the threat of extinction.
Additionally, road permeability may affect animal behavior.
Clearly, models, and especially those that incorporate scale-oriented spatial heterogeneity, that can
predict road permeability patterns should have significant contribution to reduce negative effects to both
humans and wildlife. The main objective of this study was to identify variables (determinants) that
correlate to patterns of road permeability (e.g. road-kills and passage-use) at different spatial scales. I
identified explanatory variables for road-kill and passage-use patterns by comparing various models using
'R' statistical program and by model selection. The models are based on ecological mechanisms and
composed of variables describing landscape heterogeneity, road attributes and passage characteristics. I
surveyed three roads located in northern Judea lowlands, Israel (total road length of ~45km). Road-kill
surveys were conducted by low-speed driving at dawn, 5-10 days per month. Passage-use patterns (use
rate and predation pressure) were documented by IR remote sensing cameras and an artificial nest
predation experiment. Landscape heterogeneity was incorporated into GIS, using orthophotos and field
measurements. I analyzed data using geostatistics applications and multivariate statistical tools to
correlate road attributes, their spatial context and species characteristics with observations of road-kills
and passage-use patterns. I documented 718 road-kills, with highest number of observations at the more
intense road. Ripley's k statistical analysis revealed that road-kill locations were spatially aggregated in
all three roads. However, the spatial range in which aggregations were found differed between the two
less intense roads and the more intense one, suggesting that the latter imposed a more profound
fragmentation on the landscape. Additionally, I found significant effects of the spatial context of specific
I
road-sections and passages, with determinants from different scales affecting road-kill and passage-use
patterns. I found that factors describing the spatial context of road-sections at a local scale, such as
illumination, distance to nearest crossing passage and the road-landscape conjunction, clearly affected
road-kill rates. Similarly, factors that related to the local-scale spatial context of crossing passages (such
as plant cover and human use levels) evidently affected passage-use patterns. At a broader scale, I found
that factors describing the spatial context of road-sections and passages at a landscape scale (i.e. the
matrix of land-uses), affected the related road permeability patterns (i.e. road-kills and passage-use). I
argue that this matrix of land-uses reflects the distribution of resources, and as such determines species
abundance. Moreover, the matrix of land-uses determines main movement paths, as different land-uses
serve as the attracting or deterring forces for wildlife movement.
In conclusion, the results of this study suggest that road permeability patterns relate to two main
processes: 1) at the landscape scale, wildlife activity (the flux of movement) is determined by the roadsection accessibility and its spatial context, and 2) at the local scale, spatial determinants affect the
decision of an individual regarding road-crossing options, thus influencing the result of crossing events
within the specific road-section. Consequently, these processes need to be considered when applying
mitigation programs for road-induced fragmentation.
II
Chapter 1 - INTRODUCTION
Civilizations have a long history of interaction with the landscape, exploiting natural resources and
settling human communities throughout the world. This has promoted fragmentation of the landscape on
a vast scale. A typical landscape can be considered as a mosaic of habitat patches and their
interconnections. Consequently, landscape fragmentation is the result of transforming large habitat
patches into smaller, more isolated fragments (Turner 1989). This process is very evident in urbanized or
otherwise intensively used regions, where fragmentation is the product of the connection of developed
areas via linear infrastructure, such as roads and railroads (Saunders et al. 1991, Forman and Alexander
1998, van der Ree et al. 2011). Many landscape ecological studies focus on the effects that spatial
patterning and changes in landscape structure (e.g. landscape fragmentation) have on the distribution,
movement, and persistence of species. Moreover, the habitat loss and fragmentation associated with
human land use in many regions is well described in landscape ecology and conservation biology
(Saunders et al. 1991, Heilman et al. 2002, Riitters et al. 2004, Turner 2005b). Given that landscape
connectivity plays an important role in species persistence, landscape fragmentation is now recognized as
one of the major threats to wildlife and biodiversity (Saunders et al. 1991, Forman 1995).
A major human agent of landscape fragmentation is the ever-increasing network of roads worldwide
(Forman et al. 2002), which have been shown to be harmful to various faunal groups, including
invertebrates (Haskell 2000), amphibians (Carr and Fahrig 2001), reptiles (Shriver et al. 2004), birds
(Kuitunen et al. 1998) and mammals (Philcox et al. 1999). Although vital to our economy and modernlife culture, roads impose significant ecological costs, including habitat loss and increased betweenpopulation isolation (Spellerberg 1998, Jackson 1999, Trombulak and Frissell 2000), wildlife mortality
(Malo et al. 2004, Mata et al. 2009), reduced access to resources (Forman et al. 2002, Clevenger et al.
2003), population isolation (Oehler and Litvaitis 1996, Forman and Alexander 1998, Spellerberg 1998,
Trombulak and Frissell 2000), air, ground and water pollution (Trombulak and Frissell 2000, Forman et
al. 2003), and disruption of ecological processes (Forman and Alexander 1998, Bolger et al. 2001,
Forman et al. 2003, Clevenger et al. 2003, van der Ree et al. 2007, van der Grift and Schippers 2013) .
Eventually, roads also promote high rates of wildlife-vehicle collisions (Hodson 1965, Oxley et al. 1974,
Ferreras et al. 1992, Philcox et al. 1999, Shriver et al. 2004, Taylor and Goldingay 2004), one of
mankind’s most visible road-related sources of impact upon wildlife. A growing body of literature in the
field of road ecology suggests that wildlife-vehicle collisions can be a major source of vertebrate
mortality and thus potentially limit wildlife populations (Aresco 2005). Malo et al. (2004) estimated that
vehicle-related wildlife mortality totals several millions vertebrates per year, concluding that roads may
pose the greatest mortality threat to wildlife.
1
Roads and traffic can act as barriers, which interfere with animal movement and reduce population
connectivity. As a result they can diminish gene flow and disrupt sink-source population dynamics,
thereby promoting inbreeding and loss of genetic diversity (Ferreras 2001). Consequently, roads might
increase the extinction risk of local populations, due to stochastic effects (van der Zande et al. 1980,
Saunders et al. 1991, Fahrig and Merrian 1994, Cooper and Walters 2002). Because traffic on roads acts
more like a filter to movement rather than an absolute barrier, this effect is commonly described as
relative road permeability: the higher the movement rate across it, the greater the permeability of the road.
It is well documented that the road permeability can alter animal assemblage composition, create patchypopulations, reduce biological diversity and increase the threat of extinction (Oehler and Litvaitis 1996,
Forman and Alexander 1998, Spellerberg 1998, Trombulak and Frissell 2000). Additionally, road
permeability may affect animal behavior. At the extreme scenario, poor road permeability even results in
complete reluctance to cross a road, leading to road avoidance effect (Forman and Alexander 1998). For
a more detailed review of road permeability effects on wildlife, see Forman and Alexander (1998),
Spellerberg (1998), Trombulak and Frissell (2000) and Forman (2003).
Road-kills are the most evident pattern of road permeability. However, although road-kills pose the
most obvious threat of roads to wildlife populations, the indirect effect of low road permeability (e.g.
barrier to wildlife movement, reduced access to resources and population isolation) may pose a significant
threat as well (Noss et al. 1996). While direct impacts such as road-kills are easy to document, indirect
effects are much more difficult to record and assess. Few studies have demonstrated that roads act as
barriers to the movement of small (Oxley et al. 1974, Barnett et al. 1978, Mader 1984) as well as large
vertebrates (Gibeau et al. 2002, Clevenger et al. 2003). The particular effects heavily depend on the
characteristics of the road as well as the surrounding habitat (Forman and Alexander 1998, Alexander et
al. 2005). Thus, low-grade road permeability is a major challenge for wildlife conservation.
It is generally accepted that it is unrealistic to aim for the complete elimination of the problem.
Practically, the goal should be to reduce the road-kill rates to socially tolerable levels while at the same
time promote safe movement across roads. Therefore, a major target for conservationists, when trying to
minimize the negative effects of roads on wildlife, is to improve road permeability to animal movements
in order to promote or establish contact between the populations on either side of the road. This
improvement of connectivity among isolated habitat patches in order to improve biotic exchange across
barriers, is usually termed ‘mitigation’ in highway design and habitat management disciplines (Swart and
Lawes 1996, Forman and Alexander 1998). Mitigation practices include fencing to prohibit access onto
the roadway, reduce the probability of wildlife-vehicle collision, and direct wildlife to safe crossing
passages for connecting habitats on either side of the highway (Jackson 1999). Practically, mitigation
practices have two main goals: reducing road-kills and promoting safe movement across roads.
2
A growing body of research has examined the design, implementation and efficacy of various
mitigation practices (Malo et al. 2004); some of these studies have questioned the effectiveness of these
mitigation practices (Corlatti et al. 2009). In fact, these studies correspond with the two main goals of
mitigation practices. The first type of research deals with road-kill patterns, primarily focusing on
quantifying numbers and locations of road-killed animals (e.g. Rosen and Lowe 1994, Bruinderink and
Hazebroek 1996, Huber et al. 1998, Haxton 2000, Shuttleworth 2001, Malo et al. 2004). The second type
of studies deals with usage rates of wildlife passages for road crossing (hereafter, passage-use patterns),
examining the usage rate of tunnels, culverts, and overpasses by wildlife (e.g. Foster and Humphrey 1995,
Yanes et al. 1995, Clevenger and Waltho 2000, Ng et al. 2004, Clevenger and Waltho 2005).
Regarding road-kill patterns, research on mammal road-kills has demonstrated that road-kills do not
occur randomly but are spatially clustered (Puglisi et al. 1974, Hubbard et al. 2000, Clevenger et al.
2001b, Joyce and Mahoney 2001). Since wildlife tends to be linked to specific habitats and adjacent land
use types, landscape spatial patterns would be expected to play an important role in determining road-kill
locations and rates (Forman and Alexander 1998). Explanatory factors vary widely between spatial scales
and taxa. Research shows that it is possible to predict the location of wildlife-vehicle collisions at two
scales -- landscape (road section) and local (within road section) -- and identify the variables associated
with road-kill patterns.
At the landscape scale, road-kill patterns show significant spatial clustering and appear to depend on
population density, species biology, habitat and landscape structure, as well as road and traffic attributes
(Clevenger et al. 2003, Malo et al. 2004); the latter include vehicle speed (Jaarsma et al. 2006), adjacent
vegetative cover (Ramp et al. 2006) and roadside topography (Clevenger et al. 2003). The influence of
traffic volume on road mortality has long been recognized (e.g., Clarke et al. 1998, Alexander et al.
2005). Up to a threshold volume of traffic, road-kills increase with increasing traffic; above that
threshold, some species become discouraged from crossing roads (Forman et al. 2003, Seiler and Sjolund
2005, Alexander et al. 2005).
Road-kill patterns are also predictable at a local scale, associated with points where animals find it
easier to cross roads and avoid the proximity of humans. Local attributes which affect road-kill patterns
include roadside embankments, proximity of humans, local plant cover and distance and type of nearby
crossing passages (Clevenger et al. 2003, Malo et al. 2004). Road-kills are rare where roadsides have
high embankments. Optimum crossing points for animals, and consequently road-kills, often concentrate
in areas where roads run at the same level with the adjacent landscape (Clevenger et al. 2003). Moreover,
animals avoid the proximity of humans at points where they cross roads, preferring to approach roads
hidden by tree and shrub cover (Bashore et al. 1985, Jaren et al. 1991, Clevenger et al. 2003, Seiler 2003).
Lastly, Malo et al. (2004) showed that the distance from crossing passages was significantly associated
3
with road-kill probabilities. Road-kill frequency was almost half near passages, and there were no
records of road-kills in any sampling point near underpasses. However, as mentioned above, different
studies show that passages have different rates of use by wildlife. These rates have also shown to be
modified by spatial attributes.
As pointed out, landscape scale spatial patterns may play an important role in determining road
crossing locations and rates (Forman and Alexander 1998). In turn, the location of a crossing passage,
particularly in relation to habitat quality, is an important attribute (Foster and Humphrey 1995, Yanes et
al. 1995, Land 1996, Clevenger and Waltho 2000, Ng et al. 2004). At this scale, landscape attributes such
as land use and resource distribution, with respect to the passage location, may also affect passage-use
rates.
Passage-use patterns may also be affected by local scale attributes. At the local scale, different
immediate locations, in which an individual needs to decide whether to cross or not, may vary greatly in
their spatial attributes. Various studies have shown that attributes such as distance to nearest crossing
structure and the structure’s dimensions, have an effect on crossing rates of wildlife (Reed et al. 1975,
Norman 1998, Cain et al. 2003). Other local scale attributes, describing road side’s heterogeneity (e.g.
plant cover and fencing) and disturbance (e.g. light, noise and human use levels), also may affect passageuse rates. For example, distance to cover was the most important landscape attribute for cougars
(negative correlation) and was a significant attribute determining passage for grizzly bears, elk and deer
(all positive correlations; Clevenger and Waltho 2005). Apparently, increased cover provides greater
protection and security for animals approaching road-crossing passages.
Moreover, predator-prey interactions may also influence rates and use patterns of wildlife roadcrossing passages (Little et al. 2002). The role of predation is most commonly associated with the direct
lethal effects of predators on prey. However the indirect effects of predation, such as the alteration of
prey behavior, may also be important when examining predator-prey interactions (Lima and Dill 1990,
Lima 1998, Loveridge et al. 2009). Behavioral responses to predation include spatial redistribution
(Ripple 2004), selection of specific habitat structure (Creel 2005, Wirsing et al. 2007), temporal and
spatial changes in activity patterns (Fenn and Macdonald 1995), increased vigilance and reduced foraging
time (Abramsky et al. 2002), or changes in group size (Lima 1995).
In turn, crossing passages have negative effects on wildlife movement, by influencing and
interrupting ecological processes such as predator-prey interactions (Little et al. 2002). Given that these
passages may serve as bottlenecks for wildlife movement, they could potentially increase prey
vulnerability to predation. Such passages are generally exposed, restricted, and are often narrow sites
(Reed et al. 1975, Yanes et al. 1995, Clevenger et al. 2001a). As such, these passages can reduce the
effectiveness of mechanisms used by prey species to avoid detection or escape predators. Respectively,
4
there is some evidence that crossing passages are used by predators to capture prey (Hunt et al. 1987,
Foster and Humphrey 1995; see review by Little et al. 2002).
This literature review suggests that the relative importance of local and landscape attributes affecting
road permeability patterns (e.g. road-kills as well as passage-use patterns) is controversial. Some studies
have argued that local attributes can be the most influential (Reed et al. 1975, Norman 1998, Cain et al.
2003). Other researchers have shown that the landscape matrix, particularly in relation to habitat quality,
is the most important attribute (Foster and Humphrey 1995, Yanes et al. 1995, Land 1996, Clevenger and
Waltho 2000, Ng et al. 2004). Most studies argue whether local design or landscape location is the most
important element influencing permeability patterns. Nevertheless, differences in how animals react, and
the resulted permeability patterns, can certainly be explained by specific species dynamics (Clevenger and
Waltho 2005). It is likely that different focuses on spatial scales and levels of organization by different
researchers produce different outcomes, and consequently ambiguous conclusions.
With respect to recent advances in science and practice of landscape ecology, it may be worthwhile to
refer and study road permeability and its resulting patterns using a scale-dependent approach. Landscape
ecology has been defined in various ways (Risser et al. 1984, Urban et al. 1987, Pickett and Cadenasso
1995, Turner 1989, Turner et al. 2001), but common to all definitions is a focus on understanding the
reciprocal interactions between spatial heterogeneity and ecological processes (Turner 2005a). Landscape
ecology also incorporates the reasons and consequences of spatial pattern at various spatial scales. In
turn, one of the key concepts in landscape ecology is scale-dependency. This concept relates to the
understanding of the relative importance of different factors (and their roles at multiple scales) in
producing landscape patterns (Turner 2005a). Essentially, the scale-dependent approach emphasizes that
particular patterns and processes may depend on the spatial scale and biological organizational level at
which they are considered. For example, while a particular process (or determinant) may well explain
patterns of individual behavior at a given spatial scale, other processes (or determinants) may explain
population distribution at a larger spatial scale. This approach represents the most contemporary progress
in spatial ecology and is highly supported by currently advanced developments of spatially-oriented
technology, such as GIS applications, remote sensing and geostatistical tools.
In agreement with applied ecology and conservation biology, despite roads being a predominant and
permanent feature on our landscapes, surprisingly few studies have specifically addressed the
permeability of roads and their role in habitat fragmentation (Alexander et al. 2005). As clearly presented
above, there is an urgent need to expand our ecological knowledge, as well as to apply scientificallybased knowledge on road permeability patterns. Importantly, applying this scientifically-based
knowledge should benefit not only biodiversity at different levels, but also save human life and injuries
through preventing animal-related car accidents (Forman and Merrick 2003, Bellis et al. 2007). In this
5
respect, it is obvious that studying road permeability should have far-reaching consequences to human
safety by pointing to locations and weak points where fatal events may have higher probability to occur.
Based on all of the above, I suggest applying a scale-dependent approach to study road permeability
patterns. More specifically, I suggest to examine specific road permeability related patterns (e.g. road-kill
patterns and passage-use patterns) and to identify the spatial attributes in different scales, which influence
these patterns. Moreover, although some observations on predator-prey interactions at road crossing
passages are mentioned in the literature, to the best of my knowledge, no study has tested this explicitly.
In my study, in light of the potential effects of crossing passages on predation pressure, I explicitly
examine possible effects of the spatial context of such passages on predation pressure in their vicinity.
Aims and main study questions
The main goal of my study is to identify the spatial attributes which influence road permeability
related patterns in the Mediterranean landscape. In order to do so, I will identify patterns of road-kills and
passage-use at different spatial scales. I will combine empirical and advanced theoretical methods to
study how local and landscape scale-dependent determinants affect road-kill and passage-use patterns of
various species.
My main study questions are:
How does landscape heterogeneity affect:
(a) Wildlife road-kill patterns;
(b) Wildlife passage-use patterns; and
(c) Predation pressure at the vicinity of road crossing passages?
Hypotheses and predictions
There are many processes operating at different scales that may influence road-kills and usage of
passages by wildlife. Each of the following hypotheses relate road-kill and passage-use patterns at a
different scale.
The random pattern hypothesis (null hypothesis) suggests that road-kill and passage-use patterns
are arbitrary. This hypothesis claims that road-kill and passage-use patterns are a result of probabilities
that are not influenced by spatial attributes or context. Accordingly, the prediction is that spatial
aggregation of road-kills or passage-use will not be found. Moreover, this hypothesis predicts that no
correlation will be found between spatial attributes and the examined patterns.
The local scale hypothesis suggests that wildlife road-kill and passage-use patterns are spatially
aggregated due to local road-section attributes. Local attributes at the crossing location such as road-side
heterogeneity (e.g. plant cover and habitat type) and disturbance (e.g. light and human use) may influence
6
the decision of an individual whether to cross at a specific location or not. These local attributes may
affect risk levels for crossing and the capability of an organism to cross the road, which in turn may
determine the patterns of road-kills and passage-use. For example, the presence or amount of vegetative
cover at passage entrances is considered an essential component for designing effective tunnels (Hunt
1987, Rodriguez 1996, Pfister 1997). The rationale is that increased cover provides greater protection and
security for animals approaching the passages. Local structure attributes have also shown to have an
effect on crossing rates of wildlife (Reed et al. 1975, Norman 1998, Cain et al. 2003, Clevenger and
Waltho 2005). For example, crossing passages with high openness (= width  height/length; Reed 1985)
ratios (i.e. short in length, high and wide), strongly influenced passage by grizzly bears, wolves, elk and
deer. Furthermore, optimal crossing points for animals, and consequently for road-kills, often concentrate
in areas where roads run at the same level with the adjacent landscape (Clevenger et al. 2003).
Additionally, animals avoid the proximity of humans at points where they cross roads, preferring to elude
contact with humans (Bashore et al. 1985, Jaren et al. 1991, Clevenger et al. 2003, Seiler 2003).
Following this hypothesis, it is predicted that spatial aggregations will be correlated to local scale
attributes and will be explained by local variables describing road-section heterogeneity. More
specifically, the prediction is that proximity of plant cover, distance to wildlife crossing passages and
passage attributes will be correlated with passage-use rates. Moreover, frequent use of passages by
humans is predicted to lower rate of use by wildlife. According to this hypothesis I also predict that road
sections which are leveled with the landscape, will exhibit higher rates of road-kills than road-sections not
evened out with the landscape. Moreover, I predict that road-kill patterns will be correlated with local
attributes such as distance to the nearest crossing structure and illumination.
The landscape scale hypothesis suggests that road-kill and passage-use patterns are influenced by
landscape attributes. Landscape ecology provides evidence that patch composition and physiognomy (the
position of the patches in the landscape) determine species movement patterns, thus affecting activity
levels at different patches (Dunning et al. 1992, Wiens et al. 1993, Forman 1995, Turner et al. 2001).
Hence, the interrelationship between patches in the landscape is significant in determining ecological
processes. The degree of isolation between patches (distance and nature of the surrounding environment)
and the similarity between adjacent patches play major roles in defining those processes (Addicott et al.
1987, Dunning et al. 1992, Turner et al. 2001). Roads are embedded within the landscape. As such the
road permeability should be related to the entire movement of an animal across the landscape.
Accordingly, this hypothesis argues that the landscape matrix influences movement pattern and funnels
movement of wildlife to specific road sections. As a result, activity levels (e.g. the flux of road-crossing
attempts) at a road section are a result of the landscape patches surrounding it (e.g. the different land
uses). In turn, these activity levels determine the patterns of wildlife road-kills and passage-use.
7
Accordingly, I predict that similarity of habitats at both sides of the road will increase road-crossing
rates. Moreover, land use analysis (e.g., sectors of natural habitat, human settlements, food resources,
etc.) may predict road-kill aggregations and passage-use pattern. Thus, according to this hypothesis, I
predict that spatial aggregations will be found at the landscape scale. Furthermore, this hypothesis
predicts that these aggregations will be explained by landscape variables. More specifically, the
prediction is that road sections between patches which are more often used by organisms (preferred
habitat, food resources etc.) will show spatial aggregations of road-kills and higher rates of passage-use.
The scale dependent hypothesis (activity/capability hypothesis) suggests that both local and
landscape variables interact to determine road-kill and passage-use patterns. This hypothesis argues that
spatial aggregations can be found at different scales. Applying a scale-dependent approach is
advantageous. At the landscape scale, species/population distribution is a process of inter-patch dynamics
(MacArthur and Wilson 1967, Levin 1970, Leibold et al. 2004). Clearly, the population distribution and
locations influence the probability of encountering a road during movement in the landscape.
Moreover, at the landscape scale, landscape matrix provides the needs for the organism. As such, the
landscape heterogeneity funnels the organism's movement into proper patches. If a road dissects the
movement path, the organism has to decide if and how to cross the road. Accordingly, landscape scale
attributes determine the core area and center of activity (i.e. certain portions within a home range that are
more frequently used by an animal; Hayne 1949, Silva and Talamoni 2004) and the main movement
vectors and as such, the specific road section at which the organism desires to cross, while local attributes
determine local capabilities for crossing the road. This suites scale hierarchy theory (Turner et al. 2001)
which argues that pattern observed at a focal scale is constrained by processes and patterns at a scale
above it, and can be explained by the components at the scale below it.
In agreement with Tischendorf and Fahrig (2000), and with respect to recent progress in landscape
ecology practice and knowledge, this hypothesis states that road permeability is determined by the
interaction between landscape scale process which determine the activity level of organisms, and the local
scale which determines crossing capability levels. Thus, road permeability relates to the scale-dependent
concept; i.e., different patterns may emerge at differing scales of investigation. Nevertheless, there is no
single natural scale at which phenomena are to be studied, because they naturally occur and operate at
multiple scales (Levin 1992). Accordingly, this hypothesis predicts that spatial aggregations of road-kill
and passage-use will be found at multiple scales. More specifically, this hypothesis predicts that
aggregations will be found at the local scale and the landscape scale, and will be correlated with local
variables describing patch heterogeneity and landscape matrix analysis, respectively.
8
Chapter 2 - DESCRIPTION OF RESEARCH
2.1 Study area
My study focuses on three specific secondary roads (inter-city, high traffic roads): Road 3, Road 44
and Road 424 (Figure 2.1.1), and their surrounding landscape -- the Soreq region at the Central Judea
Lowlands (CJL). The total road length I have surveyed and examined is approx. 45 km.
Lod
Modi'in
Karmei
Yosef
Latrun
Soreq
Figure 2.1.1: The study area and the examined roads.
2.2 Data collection
Monitoring wildlife road-kills
I have quantified and documented road-kills of any animal that was found by surveying the study area
roads. Road-kill surveys were conducted 2-4 times a week by driving slowly (30 km/h) throughout the
entire study area at dawn. I did not survey at night or in inclement weather due to poor visibility. Each
road-kill was identified, documented and recorded for location using a hand-held GPS. Surveys were
conducted during February-September 2010, February-December 2011 and January-August 2012, 137
days in total. Due to the nature of observations some identification problems occasionally arise. Thus, I
collected specimens which I was not able to identify in the field, preserved them in alcohol when
9
necessary, and brought them back to the lab for identification. In some cases identification was not
possible, so I documented the observation as unidentified (e.g. snake UI, bird UI).
A number of biases likely skewed my totals downward. For example, Slater (2002) described a large
difference between road-kills counted by car and on foot. Another factor leading to under-counting is
rapid decomposition when warm and humid. Furthermore, critically injured animals could crawl off
roads into nearby vegetation, and road clean-up crews sometimes remove carcasses. Another issue which
could bias my analyses was potential secondary road-kills (e.g. intense scavenging). First, I did not find
significant evidence for this phenomenon in the field (e.g. very closeby road-kills of predators and prey
on the same survey day). Secondly, a recent study by Teixeira et al. (2013) concludes that coincidence of
road-kill hotspot location is relatively low between large and small animals. This implies that secondary
road-kills are of low magnitude and should not significantly bias road-kill surveys. Despite these
potential biases my data allow relative comparisons and provide a representative sample of road-kills in
the Central Judea Lowlands.
Monitoring wildlife use of road-crossing passages
Monitoring animal movement within the passageways is important in determining whether the
passages are functional and what animals use them. Usage of passages was monitored by IR remote
sensing cameras (HCO ScoutGuard SG550V). The HCO ScoutGuard SG550V is a trail camera which is
designed for out-door wildlife tracking. This unit incorporates a wide detection zone with a fast trigger
and a 6 second recovery time. The IR sensors activate the camera by detecting heat and/or motion.
Figure 2.2.1: Representative road-crossing passages in the present research.
10
Monitoring predation pressure at the vicinity of road-crossing passages
In order to quantify predation pressure, I used a known practice of Artificial Nest Predation
(Soderstrom et al. 1998). Ten passages were surveyed in the northern Judea lowlands, in road numbers '3'
and '424' (5 passages in each road). Artificial nests were placed parallel to the road, at a distance of 50 m
from each other, at total range of 200 m of each side of the passage (Figure 2.2.2). The nests were placed
on the ground and were composed of a wicker basket containing two chicken eggs (see examples in
Figure 2.2.3). I visited each basket every 7-14 days and documented the presence or absence of eggs
(binomial). The frequency of absent eggs reflected the predation pressure. I replenished the nests with
new eggs on every visit.
Figure 2.2.2: Basket distribution in the vicinity of passages.
Figure 2.2.3: Wicker baskets containing two chicken eggs.
11
Environmental variables
I collected information on environmental variables related to two examined entities, roads and roadcrossing passages. I transected each examined road into consecutive sections of 100 m length, according
to preliminary results of Ripley's K statistics and knowledge of the examined study area (see below). The
Ripley's K results showed that road-kills are aggregated within this scale of observation. A transection
longer than 100 m might exceed the logical scale in which local scale attributes are reasonable. Whereas,
road-sections shorter than 100 m might raise problems with number of observations per section and might
even be biased by the GPS. As for road-crossing passages, any structure which could be used by wildlife
to cross the road, was documented. I assigned each entity (road-section or road-crossing structure) with
local and landscape attributes and incorporated all into GIS layers.
Local Scale attributes
Road-sections
I documented road-sections’ local variables by direct observations and by using GIS software tools.
First, I documented illumination by the presence or absence of artificial lighting. Secondly, I assigned
each road-section as flat or uneven, with respect to its road-surface conjunction. Lastly, I calculated the
distance from the road-section's edge to the nearest road-crossing structure. Road-sections containing a
road-crossing structure were assigned with a distance of 0. I incorporated all these variables into the GIS
software.
Road-crossing passages
I quantified various local scale variables at each structure’s vicinity by direct observations. I visually
estimated plant cover at ~10 m radius from the passage entrance. I also measured and quantified the
passages' physical dimensions and calculated the passage openness indices (= width  height/length; Reed
1985). I also quantified human activity by counts of people on foot, bike, and horseback and generated a
human use level index for each passage (e.g. an ordinal variable which reflects the level of usage of these
passages by humans). I furthermore measured the distance of open area between the verge of the nearest
land use and the entrance to the structure (in meters). Lastly, I measured the vertical distance from the
top of the structure to the road surface (hereafter: height from road surface).
Landscape scale attributes (road-sections and road-crossing passages)
I used orthophotos and spatial entities, as well as GIS methodology, to access and analyze landscape
scale heterogeneity. I characterized and mapped the different patches (relative homogeneous spatial
units) of the entire NJL landscape by explicitly identifying each existing patch. All patches were located
12
in the GIS-based matrix of the landscape, assigned a unique ID, classified according to land use (natural,
agricultural, human planted forest, big water resource etc.), and described by geographical parameters
(area and perimeter) and position (see Haining 2003, Fortin 2005).
In order to determine the effect of the landscape scale I chose variables describing the matrix of land
uses. The chosen land uses were obtained utilizing an orthophoto and GIS layers. Accordingly I
digitized the study area and assigned each polygon to one of the following land uses:
1. Dwarf-shrub Steppe (e.g. Batha; DsS) – a formation including dwarf shrubs of Mediterranean
origin (such as Sarcopoterium spinosum) and of Irano-Turanian origin (such as Noaea
mucronata and Artemisia sieberi). The main vegetation association is characterized by the
Ballota undulatae.
2. Natural Forest (NF) – Mediterranean forests composed mainly of Ceratonia siliqua and
Pistacia lentiscus. Tree association also includes Quercus calliprinos, Pistacia palaestina,
Crataegus aronia and Phillyrea latifolia.
3. Human Planted Forest (HPF) – composed of mainly Pinus halepensis, Ceratonia siliqua,
Cupressus and Olea europaea. Unlike the natural forests, the distribution of trees is highly
uniform and denser.
4. Human Settlement (HS) – mainly small-size townships (e.g. community villages and
"moshav"s) with a population size ranging from a few hundred to a few thousand (sometimes
tens of thousands) inhabitants. The major practices of these townships are agriculture and
farming.
5. Perennial Agriculture (PA) – mainly vineyards and plantations of fruits from the Citrus
genus.
6. Annual Agriculture (AA) – mainly wheat crops.
7. Abandoned fields (AF) – fields which have not been in use for at least ~10 years.
I calculated the percentage of these different land-use coverages in buffers around the examined
entities (road-sections and/or passages). I used buffers with radii of 100 m, 250 m, 500 m, 1000 m and
2000 m.
2.3 Data analysis
Road kill aggregations
Roads were transected into consecutive road-sections of 100 m in length. Each single road section
was documented and assigned with local and landscape attributes. I used Average Nearest Neighbor
Index (ANNI) to inspect for spatial aggregation. ANNI calculates a nearest neighbor index based on the
13
average distance from each feature to its nearest neighboring feature. ANNI is expressed as the ratio of
the Observed Mean Distance to the Expected Mean Distance (Ebdon 1985, Mitchell 2005). The expected
distance is the average distance between neighbors in a hypothetical random distribution. If the index is
less than 1, the pattern exhibits clustering; if the index is greater than 1, the trend is toward dispersion or
competition.
Following the ANNI analysis I used Ripley’s K-statistic to identify spatial clusters and the proper
scale in which they are found. Ripley’s K-statistic is a way to analyze the spatial pattern of incident point
data. This method describes the dispersion of data over a range of spatial scales (Ripley 1981, Cressie
1991). In order to conduct this analysis I used ArcGIS Multi-Distance Spatial Cluster Analysis tool,
which is based on Ripley's K-Function. This tool computes the average number of neighboring features
associated with each feature; neighboring features are those closer than the distance being evaluated. As
the evaluation distance increases, each feature will typically have more neighbors. If the average number
of neighbors for a particular evaluation distance is higher/larger than the average concentration of features
throughout the study area, the distribution is considered clustered at that distance (see Getis 1984, Bailey
and Gatrell 1995, Mitchell 2005).
A distinguishing feature of this method from others is that it summarizes spatial dependence (feature
clustering or feature dispersion) over a range of distances. In many feature pattern analysis studies, the
selection of an appropriate scale of analysis is required. For example, a distance threshold or distance
band is often needed for the analysis. When exploring spatial patterns at multiple distances and spatial
scales, patterns change, often reflecting the dominance of particular spatial processes at work. Ripley's K
function illustrates how the spatial clustering or dispersion of feature centroids changes when the
neighborhood size changes (Mitchell 2005).
A number of variations of Ripley's original K-Function have been suggested. Here I implement a
common transformation of the K-Function, often referred to as L(d). The transformation L(d) :
N
A ∑N
i=1 ∑j=1,j≠i k(i, j)
√
L(d) =
πN(N − 1)
where A is area, N is the number of points, d is the distance and k(i, j) is the weight, which is 1 when the
distance between i and j is less than or equal to d and 0 when the distance between i and j is greater than
d.
To assess the significance of K-values I conducted 99 simulations of the above equation based on
random distributions of points for each of the three roads. I have used these permutations to define the
confidence limits. Deviation of the observed line above the expected line and above the upper confidence
14
limit indicates that the dataset exhibits clustering at that distance. Deviation of the observed line below
the expected line and below the lower confidence limit indicates that the dataset exhibits dispersion at that
distance. Significant clustering is defined as any value of L(d) above the upper confidence limit which
deviates above of the expected line. Significant dispersion is defined as any value of L(d) below the
lower confidence limit and below the expected line.
Passage-use rates
Passage use was examined by means of rate of use. I explored this rate of use in order to identify
correlations with local and landscape attributes. I collected data on passage-use patterns by video records
of IR cameras placed at road-crossing passages. I obtained data on 25 passages, potentially used by
wildlife species. Inspection periods were of 12 hours each time, and I labeled them daytime or nighttime.
The total number of inspection periods was 292 and 146 for daytime and nighttime, respectively. The
total number of observations of wildlife crossings within the passages was 142. I categorized the species
using these passages as predators or prey. I calculated passage-use as the total number of crosses divided
by the total number of observation periods. I separated all analyses for passage-use patterns to
observations of predatory species only and observations of prey species only. This separation was
decided upon, as evidence suggests that crossing passages are used differently by predators and prey
species (Hunt et al. 1987, Foster and Humphrey 1995; see review by Little et al. 2002). In turn, the
spatial attributes, which are linked to this difference, might also have dissimilar effect on passage-use
rates of predators in oppose to those of prey. Moreover, from a technical aspect - detectability of prey
species by the IR camera significantly differs from that of predators.
Predation pressure patterns
As mentioned, I documented observations at each visit of the artificial nests. I referred presence or
absence of eggs as predation occurrence (binomial). I calculated predation occurrence as 0 or 1 for each
visit, for presence or absent of the eggs, respectively. Furthermore, I also calculated the predation
frequency in total. I considered predation frequency as the proportion of predation occurrence out of the
total number of visits for each location.
GLMs
Based on the prospective explanatory variables I constructed a series of biologically-meaningful
General Linear Models (GLMs) for each road permeability related pattern (i.e., road-kills, passage-use
and predation in the vicinity of passages). I used different link functions with respect to the nature of the
examined dependent variable. For road-kills, which are count variables, I used the Poisson distribution
15
link function. For passage use rates I used the Gaussian distribution for linear models. For the predation
pressure related models I used the logistic regression link function, as it is a binomial variable.
I also considered parameter estimates averaged over the entire model set, proportional to the support
that each model receives (Anderson et al. 2000). This approach accounts for uncertainty in model
selection and thus leads to appropriately broader confidence intervals than would be obtained by relying
only on the single, best-supported model.
Model selection and Importance index
Anderson and Burnham (2002) present model selection and multi-model inference as mechanisms to
allow ranking and weighting of models as well as selection of the ‘best’ model from a predefined a priori
set. Akaike's Information Criterion (AIC), or one of a number of related information criteria (Anderson
and Burnham 2002), is used to rank a series of models applied to a particular dataset. In this study I used
AIC with second order correction (corrected for small sample size; AICc). I then used the difference in
AICc for each model in the set compared with the minimum AICc (for the best model) in order to
calculate Akaike weights (Buckland et al. 1997). These can be interpreted as the probability that each
model is the best model, given the data and set of models (Anderson and Burnham 2002). Akaike
weights can also be used to evaluate the relative contribution of different variables in the set of models.
This relative contribution is referred to as importance index. The weights are summed for all models that
contain a given variable to give an indication of the relative importance (hence the name; importance
index) of variables across the set of models. All of my analyses which are reported in this study, were
implemented in the R programming language (R Development Core Team 2011), using functions from
the R packages presented in Table 2.3.1.
16
Table 2.3.1: R packages used in this study.
R package name Description
Author*
AICcmodavg
Model selection and multimodel inference based on (Q)AIC(c)
Mazerolle 2013
BiodiversityR
GUI for biodiversity and community ecology analysis
Kindt 2011
combinat
Routines for combinatorics
Chasalow 2010
cshapes
Package for CShapes, a GIS dataset of country boundaries
(1946-2008). Includes functions for data extraction and the
computation of distances matrices and -lists
entropy
faraway
Entropy and mutual information estimation
Weidmann and
Skrede 2012
Hausser and
Strimmer 2012
Functions and datasets for books by Julian Faraway. Books are
"Practical Regression and ANOVA in R" on CRAN, "Linear
Models with R" published in August 2004 by CRC press and
Faraway 2011
"Extending the Linear Model with R" published by CRC press
in December 2005
gregmisc
The original gregmisc bundle is a repository for a variety of
useful functions. The gregmisc bundle was recently split into a
set of more focused packages: gdata, gmodels, gplots, gtools.
The purpose of this 'new' gregmisc is to provide an easy way to
Warnes 2011
access the original combined functionality. To this end, it
simply depends on all of the new packages so that these will
installed/loaded when this package is installed/loaded
lme4
maptools
Fit linear and generalized linear mixed-effects models
Bates et al.
2012
Set of tools for manipulating and reading geographic data, in
particular ESRI shapefiles; C code used from shapelib. Includes
binary access to GSHHS shoreline files. The package also
Lewin-Koh and
provides interface wrappers for exchanging spatial objects
Bivand 2012
with packages such as PBSmapping, spatstat, maps, RArcInfo,
Stata tmap, WinBUGS, Mondrian, and others
MASS
Functions and datasets to support Venables and Ripley, 'Modern
Ripley et al.
Applied Statistics with S' (4th edition, 2002)
2012
17
plotrix
Lemon at el.
Various plotting functions
2012
Interface to Geometry Engine - Open Source (GEOS) using the
Bivand et al.
C API for topology operations on geometries
2012
shapefiles
Functions to read and write ESRI shapefiles
Stabler 2006
sp
A package that provides classes and methods for spatial data.
rgeos
The classes document where the spatial location information
resides, for 2D or 3D data. Utility functions are provided, e.g.
for plotting data as maps, spatial selection, as well as methods
Pebezma and
Bivand 2012
for retrieving coordinates, for subsetting, print, summary, etc.
spdep
Spatial dependence: weighting schemes, statistics and models
Bivand 2012
Rowlingson
splancs
Spatial and Space-Time Point Pattern Analysis Functions
and Diggle
2012
vegan

Ordination methods, diversity analysis and other functions for
Oksanen et al.
community and vegetation ecologists
2012
Author stands for the developer of the package as mentioned at the R help description file of
the package.
The appliance of the scale-dependent approach
As previously mentioned, one of the innovations of this study is the applying of a scale-dependent
approach. With respect to this approach, and according to my hypotheses, I constructed the following
regime for my analyses:
1. Analyzing the pattern itself and examining clustering and differences from random patterns;
2. Identifying effects of local scale variables;
3. Identifying effects of landscape scale variables:
a. Of all variables at each one of the different radii of observation – using information
criteria statistics I calculated an ‘importance index’ indicating the relative
contribution of different variables in each set of models.
b. Of each variable across radii of observation – I examined the change in effect as
observation radius changes. For each specific variable I examined the change in the
importance index values and the correlation's direction according to their related
coefficients, across observation radii.
18
c. Of variables from different observation radii – I identified the radius of observation
in which each specific variable has the most profound effect and designed a new set
of models constructed by all the combinations of these variables (with no
interactions). I examined each variable's importance index value and coefficient.
4. Identifying multi-scale effects – I analyzed the relative joint contribution of local- and
landscape-scale variables, i.e. multi-scale effects. I selected the variables which exhibit the
highest importance index values in their related scale, based on the previous analyses (e.g. for
each scale separately; steps 2 and 3).
19
Chapter 3 – RESULTS
3.1 Road-kill patterns
Road-kills are perhaps the most visible pattern of road-induced fragmentation. In order to examine
the effect of landscape heterogeneity on patterns of road-kills, I first explored and identified the spatial
pattern of the road-kill locations. I collected and documented 708 observations of road-kill events during
three sampling seasons (2010, 2011 and 2012; a total of 134 days; see Chapter 2). Figure 3.1.1 presents
the number and relative percentage of road-kill observations of each taxonomic class for each road. The
different roads did not differ significantly in the distribution of road-kills by classes. I found that most
road-kills are of mammals (approx. 40%). Counter-intuitively, high numbers of road-kills were of birds,
with 256 recorded events, ~35% of all observations. Lastly, reptiles represented ~25% of the total
number of road-kill observations.
282
256
300
Number of road-kills
Amphibians
Birds
Mammals
Reptiles
250
200
167
136
112
150
84
100
60
52
68
61
86
46
50
3
3
0
'3'
'44'
'424'
Grand Total
Road
Figure 3.1.1: Road-kill events by roads of each one of the examined taxonomic classes. Bars present
the count and table presents percentages.
Table 3.1.1 presents the list of species that I found road-killed and their numbers per road. Due to
identification difficulties (see Chapter 2) some observations were marked as unidentified (UI). When
possible I assigned unidentified birds as small, medium or large.
20
Table 3.1.1: List of road-killed species and number of observations at each road.
Road
Class and Species
Amphibia
Bufo viridis
Avians
Alaudidae
Asio
Asio otus
Bird UI
Buteo rufinus
Cinnyris oseus
Clamator glandarius
Columba
Corvus
Corvus cornix
Coturnix
Egretta garzetta
Erithacus rubecula
Falco tinnunculus
Gallus gallus
domesticus
Garrulus
Halcyon smyrnensis
Hirundo daurica
Large Bird
Lanius
Lanius nubicus
Medium Bird
Otus scops
Passer domesticus
Passer hispaniolensis
Prinia gracilis
Pycnonotus
Small Bird
Streptopelia
senegalensis
Sylvia
Turdus merula
Tyto alba
Upupa epops
Vanellus spinosus
Mammals
'3'
3
'44'
'424'
3
136
6
4
19
20
3
1
1
1
1
10
3
52
68
256
3
3
8
5
29
1
1
1
32
6
2
4
1
2
17
2
2
1
1
8
1
1
1
9
2
1
2
5
1
2
3
1
11
18
1
1
1
Total
3
11
2
2
5
1
2
1
1
1
4
2
32
1
2
1
1
1
6
1
14
1
5
7
8
3
6
1
10
4
1
8
1
6
1
3
4
6
1
3
18
5
16
112
84
86
282
3
1
9
1
1
7
5
9
1
1
1
2
1
5
21
Canis aureus
Erinaceus
Felis catus
Gazella gazella gazella
Herpestes ichneumon
Hystrix indica
Meles meles
Pipistrellus kuhlii
Rattus norvegicus
Rattus rattus
Rousettus aegyptiacus
Spalax ehrenbergi
Vulpes vulpes
Reptiles
Chalcides ocellatus
Chamaeleo
chamaeleon
Coluber jugularis
Coluber najadum
Coluber nummifer
Eryx jaculus
Laudakia stellio
Natrix tessellata
Snake UI
Telescopus fallax
Testudo graeca
terrestris
Vipera palaestinae
Eumeces schneideri
Grand Total
21
7
17
31
8
11
14
1
10
2
12
22
18
12
2
1
19
3
1
22
41
40
49
1
53
4
1
1
44
4
3
2
39
13
14
3
1
2
2
12
60
61
46
167
1
1
1
1
8
8
17
24
1
1
10
4
2
2
3
13
2
7
1
6
10
5
5
2
1
5
44
4
1
21
19
4
3
14
1
5
6
4
8
10
6
10
28
6
311
197
200
708
Next, I examined the spatial distribution of road-kill locations and the potential effect of spatial
context on road-kill numbers and aggregations. I conducted this analysis for the three main taxonomic
classes found (birds, mammals and reptiles). In the next sub-sections I present my results for each one of
these taxonomic classes.
22
3.1.1 Birds
Spatial distribution
I analyzed the spatial distribution of bird road-kill locations using Average Nearest Neighbor Index
(ANNI). I conducted the analysis for each one of the surveyed roads. I found (Table 3.1.1.1) that the bird
road-kill patterns were aggregated, for all three examined roads. The ANNI for all the examined roads
differed significantly from the expected value for random or over-dispersed patterns (N=136, N=52,
N=68, for roads 3, 44 and 424, respectively; P<0.0001 for all roads). Therefore, road-kills of birds were
spatially clustered in all the examined roads.
Table 3.1.1.1: Summary of the average nearest neighbor analysis for bird road-kills.
Birds
Observed Mean Distance
Expected Mean Distance
Nearest Neighbor Ratio
z-score
p-value
Road 3
68.605798
148.735228
0.461261
-12.019283
<0.0001
Road 44
99.332732
149.156915
0.665961
-4.608179
<0.0001
Road 424
75.854756
151.875840
0.499452
-7.896428
<0.0001
In order to determine whether road-kills of birds exhibit statistically significant clustering or
dispersion over a range of distances I conducted Ripley's K-statistics. When exploring spatial patterns at
multiple distances and spatial scales, patterns change, often reflecting the dominance of particular spatial
processes at work. Ripley's K-function illustrates how the spatial clustering or dispersion of feature
centroids changes when the neighborhood size changes. Although road-kills were spatially aggregated in
all three roads, the examined roads differed in the spatial range at which these aggregations existed
(Figure 3.1.1.1). I found that road-kills of birds were spatially aggregated up to an observation range of
~400 m in roads ‘44’ and ‘424’, in comparison with an observation range of up to 750 m in road ‘3’. This
implies that road ‘3’ imposes a more profound fragmentation on the landscape, funneling movement of
birds to more specific locations.
23
Road 44
Road 424
~ 450 m
~ 300 m
Road 3
Road 3
~ 750 m
------ Expected K
------ Observed K
------ Confidence Env.
Figure 3.1.1.1: Spatial analysis of road-kill aggregations, utilizing a common transformation of
Ripley's k-function where the expected result with a random set of points is equal to the input
distance.
Based on the clustering results, I further examined the effect of the landscape heterogeneity on these
aggregation patterns. In accordance with the scale-dependence approach, the following sections describe
my analyses in two main spatial scales -- local and landscape -- and in a multi-scale manner (i.e. including
variables from both scales).
24
Local scale effects
First, I examined the effect of artificial illumination and the relative height of the road from its
surroundings (road-surface conjunction) on numbers of road-kills of birds. Due to the fact that the entire
length of road '3' is illuminated, I conducted the illumination analysis for roads '44' and '424' only.
Likewise, because the entire lengths of roads '44' and '424' are on the same level as their surroundings, the
height analysis was conducted for road '3' only.
I did not find a significant effect of illumination on numbers of bird road-kills (Figure 3.1.1.2; p=0.4).
Figure 3.1.1.3 presents the effect of the relative level of the road compared to its surroundings on roadkills. Significantly higher numbers of bird road-kills were found on road-sections that were at the same
Number of road-kills
level as their surroundings, in comparison with sections uneven with their surroundings. (p<<0.01).
p=0.4
p = 0.4
Illumination:
Absence = 0 Presence = 1
Number of road-kills
Figure 3.1.1.2: The effect of illumination on amounts of bird road-kills on roads 44 and 424
p<<0.01
Road-surface conjunction:
Even = 0
Uneven = 1
Figure 3.1.1.3: The effect of the relative level of the road compared to its surroundings (e.g. roadsurface conjunction) on bird road-kills.
I also examined the effect of the distance to the nearest road-crossing passage on road-kill numbers.
Due to the heteroscedasticity of the variance I grouped the distances to the nearest crossing passage into
25
100 m blocks (e.g. 0-100 m, 101-200 m, etc.). I found a significant negative correlation (logarithmic
decline) between the number of bird road-kills and the distance to the nearest road-crossing passage
(p<<0.001; Figure 3.1.1.4). This means that surprisingly more road-kills of birds occur close-by to
passages rather than far away from them.
120
Number of road-kills
100
80
60
y = -29.95ln(x) + 74.202
R² = 0.855
p<<0.001
40
20
0
Distance to nearest crossing passage (m)
Figure 3.1.1.4: The effect of the distance to the nearest road-crossing passage on road-kill numbers
of birds.
Landscape scale effects
Following my hypothesis, I examined the effect of landscape scale variables on numbers of bird roadkills. First, I calculated the cover percentage of different land-uses in increasing radii of observation (i.e.
100, 250, 500, 1000 and 2000 meters) around road-sections. In order to identify patterns of the variables'
effect at each observation radius, I examined the effect of these land-use covers at each observation radius
on the observed bird road-kill numbers. Using information criteria statistics I calculated an ‘importance
index’, indicating the relative contribution of different variables in the set of models (see Chapter 2).
Table 3.1.1.2 shows a summary of the importance index values of the examined landscape scale variables
at each observation radius.
26
Table 3.1.1.2: Summary of the sets of models for the 5 landscape observation radii. Table shows
importance index values of the landscape scale variables on road-kill numbers and the AICc weight of
the best model in each set. The importance index values of variables which were included in the best
model are marked with *. The trend of the variables' effects are color-coded with red and blue for
positive and negative correlation, respectively (based on their respected coefficients). The examined
landscape variables are: PA- Perennial Agriculture, AA- Annual Agriculture, AF- Abandon Fields,
HPF- Human Planted Forest, HS- Human Settlements, DsS- Dwarf-shrub Steppe and NF- Natural
Forest.
Class
Birds
Radius
PA
AA
AF
*
HPF
HS
DsS
*
*
NF
Best model AICc wt.
*
100
0.52
0.48
0.35
0.39
0.52
0.97
0.98
0.10
250
0.72*
0.79*
0.60*
0.58*
0.57*
0.88*
1.00*
0.26
500
*
*
*
*
*
*
*
0.38
*
0.08
*
0.08
1000
2000
0.72
0.83
0.45
0.43
0.47
*
0.68
0.87
*
0.95
*
0.87
0.73
0.42
0.36
0.69
0.35
0.37
0.84
*
0.65
*
0.52
1.00
0.98
0.67
At small radii within the landscape scale (100 m and 250 m), bird road-kill numbers were best
explained by percentage of DsS and NF covers. At intermediate observation radius (500 m) road-kills
were best explained by the combination of all relative covers and were likely affected by the total spatial
heterogeneity. At the largest observation radii (1000 and 2000 m), bird road-kill numbers were best
explained by the land-cover of AF and NF.
Particular variables that were most important at certain observation radii, were of certain importance
in other radii of observation. Therefore, in order to conduct a robust analysis for the landscape scale
variables and to accurately examine their effect on bird road-kill numbers, I first examined the change in
effect as function of the observation radius. For each specific variable I examined the importance index
values across all observation radii (Figure 3.1.1.5) and the correlation's direction based on their related
coefficients (Table 3.1.1.2). Second, I identified the radius of observation at which each specific variable
had the most profound effect and designed a new set of models constructed by all the combinations of
these variables (with no interactions). I examined each variable's importance index value and coefficient
(Table 3.1.1.3).
My first examination (Figure 3.1.1.5) showed that importance index values of only two variables
(DsS and NF) were significantly correlated with radii of observation. Both correlations were negative
(p=0.005 and 0.04, R2=0.95 and 0.81, for DsS and NF, respectively). This implies that the effect of DsS
and NF landcovers on road-kills of birds decreases as distance from the road-section increases.
27
PA
Importance index value
1.00
AF
AA
1.00
1.00
0.90
0.90
0.90
0.80
0.80
0.80
0.70
0.70
0.70
0.60
0.60
0.60
0.50
0.50
0.50
0.40
0.40
0.40
0.30
0.30
0.30
0.20
0.20
0.20
0.10
0.10
0.10
0.00
0.00
0
500
1000
1500
2000
0.00
0
500
1000
HPF
1500
1.00
0.90
0.90
0.80
0.80
0.70
0.70
0.60
0.60
0.50
0.50
0.40
0.40
0.30
0.30
0.20
0.20
0.10
0.10
0.00
1000
1500
2000
y = -0.0002x + 0.9482
R² = 0.9503
p = 0.005
1500
2000
0
500
1000
1000
1500
2000
NF
0.00
500
500
DsS
1.00
0
0
2000
1500
2000
1.10
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
y = -0.0002x + 1.0525
R² = 0.8091
p = 0.04
0
500
1000
1500
2000
HS
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0
500
1000
Radius of observation (m)
Figure 3.1.1.5: The effect of landscape scale variables (cover percentage of different land-uses) on
bird road-kills. Graph shows importance index values as a function of observation radius, for each
examined variable. The examined landscape variables are: PA- Perennial Agriculture, AA- Annual
Agriculture, AF- Abandon Fields, HPF- Human Planted Forest, HS- Human Settlements, DsSDwarf-shrub Steppe and NF- Natural Forest.
Additionally, I found that the change in the correlation's direction (based on related coefficients)
across observation radii differed between the examined variables (Table 3.1.1.2). Some variables
exhibited a consistent effect on bird road-kill numbers, such as the negative correlation with PA, AF and
NF at all observation radii. Other variables -- AA, HS and DsS – exhibited a positive correlation in small
and large observation radii and a negative correlation in intermediate observation radii. However, when
considering their importance index values, I found that AA and HS were of the highest importance at an
observation radius of 500 m, whereas the highest importance value of DsS was at 100 m. Thus, the
results suggest that the negative correlations of AA and HS have a more profound effect on bird road-kill
28
numbers, whereas the positive correlation is more profound for DsS. Lastly, HPF correlated differently
with road-kill numbers, from negative to positive at the largest observation radius. Keeping in mind that
HPF exhibited the highest importance index at the observation radius of 500 m, implying that the negative
correlation has higher weight on road-kill numbers than the positive one.
Next, as mentioned above, I selected each variable at the radius at which it exhibited the highest
importance index value and designed a new set of models based on all the combinations of these variables
(with no interactions).
Table 3.1.1.3: Summary of the importance index values and coefficient estimates of the variables'
effect on bird road-kill numbers at the landscape scale. Variables included in the models are at the
radius in which they previously exhibited highest importance index values. PA- Perennial
Agriculture, AA- Annual Agriculture, AF- Abandon Fields, HPF- Human Planted Forest, HS- Human
Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Class
Variable
Radius of observation
Importance
Coef_est
Birds
DsS
100
1.00
2.40
NF
500
0.99
-4.76
AA
500
0.79
0.95
AF
1000
0.72
-2.67
PA
500
0.40
-0.76
HS
500
0.34
-0.43
HPF
500
0.34
-0.57
Most variables exhibited the highest importance index at an observation radius of 500 m (5 out of 7).
From my analysis of the entire model set, I found that the most important variables for predicting roadkill numbers of birds were DsS and NF, with importance index values of 1 and 0.99, respectively. I also
found moderate importance values for cover percentages of AA and AF (0.79 and 0.72, respectively).
Furthermore, DsS and AA were positively correlated with road-kill numbers, while NF and AF exhibited
a negative correlation. This means that higher cover percentage of DsS and AA around road-sections
make them more prone to road-kills of birds. On the other hand, lower cover percentage of NF and AF
around road-sections also makes them more prone to road-kills of birds.
Multi-scale effects
In accordance with the scale-dependent approach, I aimed for a better understanding of the overall
effects of different factors from different scales on bird road-kill numbers. Thus, I analyzed the relative
29
joint contribution of local- and landscape-scale variables, i.e. multi-scale effects. I generated and
analyzed 128 models composed of 7 variables from different scales. I selected the variables based on the
analyses I previously conducted for each scale separately. These variables exhibited the highest
importance index values in their related scale (see above). The local-scale variables included
illumination, road-surface conjunction and distance to the nearest passage. The landscape-scale variables
included land-use covers of DsS at observation radius of 100 m, AA at observation radius of 500 m, NF at
observation radius of 500 m and AF at observation radius of 1000 m.
It appears that landscape-scale variables were of higher importance than local-scale variables in
determining numbers of bird road-kills (Table 3.1.1.4). In contrast to the analysis conducted for the local
scale alone, this analysis revealed that bird road-kill numbers were also profoundly affected by the
presence of illumination (importance index = 0.93). Moreover, the distance to the nearest passage was
also of high importance in influencing road-kill amounts of birds (importance index = 0.84).
Table 3.1.1.4: Summary of the multi-scale analysis for bird road-kill numbers. Table shows the
relative importance and correlation trend of the examined variables. DsS- Dwarf-shrub Steppe, AAAnnual Agriculture, NF- Natural Forest and AF- Abandon Fields.
Variable
Scale
Illumination
Road-surface conjunction
Distance to nearest passage
DsS
AA
NF
AF
Local
Local
Local
Landscape
Landscape
Landscape
Landscape
Radius of
observation (m)
NA
NA
NA
100
500
500
1000
30
Importance
0.93
0.3
0.84
1
0.97
0.98
0.6
Correlation
+
+
+
-
3.1.2 Mammals
Spatial distribution
Similar to the analysis I conducted for road-kills of birds, I first analyzed the spatial distribution of
mammal road-kill locations using Average Nearest Neighbor Index (ANNI). Similarly, I conducted the
analysis for each one of the surveyed roads. Results are presented in Table 3.1.2.1, showing that road-kill
patterns of mammals were aggregated, in all 3 examined roads. I found that ANNIs for all the examined
roads significantly differed from the expected value for random or over-dispersed patterns (N=112, N=84,
N=86, for roads 3, 44 and 424, respectively; p<0.0001 for all roads). Thus, I can confidently conclude
that road-kills of mammals are spatially clustered in all the examined roads.
Table 3.1.2.1: Summary of the average nearest neighbor analysis for mammal road-kill patterns.
Mammals
Observed Mean Distance
Expected Mean Distance
Nearest Neighbor Ratio
z-score
p-value
Road 3
70.659101
164.193222
0.430341
-11.533330
<0.0001
Road 44
70.235273
133.706804
0.525293
-8.323309
<0.0001
Road 424
74.941090
129.899586
0.576916
-7.505976
<0.0001
Next, I conducted a Ripley's K-statistics analysis. As previously mentioned, this analysis determines
whether features (incident point data), or the values associated with features, exhibit statistically
significant clustering or dispersion over a range of distances. Similar to the results of bird road-kills, I
found that road-kills of mammals were spatially aggregated in all three roads. Likewise, the spatial range
in which these aggregations were found differed between the examined roads (Figure 3.1.2.1). I found
that road-kills were spatially aggregated up to an observation range of ~650 m in road ‘44’ and ~500 m in
road ‘424’. In road ‘3’, my results showed aggregations of road-kills up to an observation range of ~1000
m. Again, this implies that road ‘3’ imposes a more profound fragmentation on the landscape.
31
Road 44
Road 424
~ 500 m
~ 650 m
Road 3
~ 1000 m
------ Expected K
------ Observed K
------ Confidence Env.
Figure 3.1.2.1: Spatial analysis of road-kill aggregations, utilizing a common transformation of
Ripley's k-function where the expected result with a random set of points is equal to the input
distance.
According to my clustering results, I next examined the effect of the landscape heterogeneity on these
aggregation patterns. In accordance with the scale-dependence approach, and with respect to my
hypotheses, the following sub-sections describe my analyses in two main spatial scales -- local and
landscape -- and in a multi-scale manner (i.e. including variables from both scales).
32
Local scale effects
First, I examined the effect of artificial illumination and the relative height of the road from its
surroundings (road-surface conjunction) on mammal road-kill numbers. As previously mentioned, due to
the fact that road '3' is entirely illuminated, I conducted the analysis for roads '44' and '424' only.
Likewise, because roads '44' and '424' are on the same level as their surroundings for their entire length, I
conducted this analysis for road '3' only.
I found a marginally significant (p=0.07) effect of illumination on mammal road-kills. Higher
numbers of mammal road-kills were found on illuminated road-sections, in comparison with dark sections
Number of road-kills
mammal road-kill numbers (Figure 3.1.2.2).
p = 0.07
Illumination:
Absence = 0 Presence = 1
Number of road-kills
Figure 3.1.2.2: The effect of illumination on amounts of mammals' road-kills.
p << 0.01
Road-surface conjunction:
Even = 0
Uneven = 1
Figure 3.1.2.3: The effect the relative level of the road compared to its surroundings (i.e. roadsurface conjunction) on mammals road-kill numbers.
Figure 3.1.2.3 presents the effect of the relative level of the road compared to its surroundings (i.e.
road-surface conjunction) on mammal road-kill numbers. Similar to the results for bird road-kills, higher
numbers of mammal road-kills were found in road-sections that were at the same level as their
surroundings, in comparison with un-leveled sections (p<<0.01).
33
Next, I examined the effect of the distance to the nearest road-crossing passage on mammal roadkills. Due to the heteroscedasticity of the variance I grouped the distances to the nearest crossing passage
to 100 m blocks (e.g. 0-100 m, 101-200 m, etc.). The distance to the nearest road-crossing passage was
found to be significantly correlated with mammal road-kills, exhibiting a logarithmic decline (p<<0.001;
Figure 3.1.2.4). Like the results for bird road-kills, more road-kills of mammals occur close-by to
passages rather than far away from them.
120
Number of road0kills
100
80
60
y = -31.22ln(x) + 78.183
R² = 0.849
p<<0.001
40
20
0
Distance to nearest crossing passage (m)
Figure 3.1.2.4: The effect of the distance to the nearest road-crossing passage on mammal road-kill
numbers.
Landscape scale effects
To examine the effect of landscape scale variables on mammal road-kill numbers, I first calculated
the cover percentage of different land-uses in increasing radii of observation (e.g. 100, 250, 500, 1000 and
2000 meters) around road-sections. In order to identify patterns of the variables' effect at each
observation radius, I examined the effect of these covers at each observation radius on the observed
amount of mammal road-kills. I used information criteria statistics to calculate the ‘importance index’,
which indicates the relative contribution of different variables in a set of models (see Chapter 2). Table
3.1.2.2 shows a summary of the importance indices of the examined landscape scale variables at each
observation radius.
34
Table 3.1.2.2: Summary of the sets of models for the 5 landscape observation radii. Table shows
importance index values of the landscape scale variables on road-kill numbers and the AICc weight of
the best model in each set. The importance index values of variables which were included in the best
model are marked with *. The trend of the variables' estimated coefficients are color coded with red
and blue for positive and negative correlation, respectively. The examined landscape variables are:
PA- Perennial Agriculture, AA- Annual Agriculture, AF- Abandon Fields, HPF- Human Planted
Forest, HS- Human Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Class
Mammals
Buffer_r
PA
AA
AF
HPF
*
HS
DsS
*
NF
Best model AICc wt.
*
100
0.69
0.67
0.89
0.45
0.63
0.41
0.90
0.13
250
0.81*
0.67*
0.73
0.58*
0.73*
0.49*
0.81
0.10
500
*
*
*
0.65
*
0.69
*
0.87
*
*
0.14
0.90
0.67
0.59
0.79
1000
NA
NA
NA
NA
NA
NA
NA
NA
2000
0.44
0.76*
0.39
0.34
0.39
0.74*
0.98*
0.17
The set of models for examining the effect of land-use covers at an observation radius of 1000 m on
mammal road-kills failed to meet the statistical requirements (e.g. distribution and autocorrelation of
variables). As such, results for this observation radius are not available. As for the rest of the model
sets, results showed that at an observation radius of 100 m the most important variables were land-use
covers of NF and AF (0.9 and 0.87, respectively). At an observation radius of 250 m there is also high
importance for the covers HS and PA (0.73 and 0.81, respectively). At this radius there was also a
moderate importance for the cover percentage of AA. At 500 m, PA, HS and NF were of high
importance, while all the rest were moderately important. This implies that at this observation radius (500
m) heterogeneity itself plays an important role. At the largest observation radius (2000 m) mammal roadkills were best explained by the relative cover of NF (0.98) with a moderate importance to the covers of
AA and DsS (0.76 and 0.74, respectively).
I further compared the change in variable's importance index values across observation radii for each
examined land-use cover (Figure 3.1.2.5). I performed this comparison in order to identify the radius of
observation at which each specific variable had the most profound effect and to examine trends in effect
as observation radius changed. As I previously showed for bird road-kills, I found different trends for the
different land-use covers. I found that the importance of AF decreased as radius of observation increased
(p=0.05, R2=0.9). Contrarily, importance index values of percentage of AA cover was positively
correlated with observation radius (p=0.02, R2=0.96). Lastly, importance index values of DsS and HS
showed a concave (n-shape) trend with the radius of observation (p=0.02 and 0.03, R2=0.99 and 0.99,
respectively).
35
PA
Importance index value
1
AA
1
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
0
500
1000
1500
2000
0.5
0.4
0.3
0
0
500
1000
1500
2000
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
500
1500
2000
0
500
1000
1000
1500
2000
1500
2000
NF
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
y = -2E-07x2 + 0.0005x + 0.36
R² = 0.99
p = 0.02
0
1000
0
DsS
1
500
y = -0.0002x + 0.82
R² = 0.9
p = 0.05
0.6
y = 5E-05x + 0.65
R² = 0.96
p = 0.02
HPF
1
0
AF
1
1500
2000
0
500
1000
HS
1
0.9
0.8
0.7
0.6
0.5
0.4
y = -5E-07x2 + 0.0009x + 0.54
R² = 0.99
p = 0.03
0.3
0.2
0.1
0
0
500
1000
1500
2000
Radius of observation (m)
Figure 3.1.2.5: The effect of landscape scale variables (cover percentage of different land-uses) on
mammal road-kills. Graphs show importance index values as a function of observation radius, for
each examined variable. The examined landscape variables are: PA- Perennial Agriculture, AAAnnual Agriculture, AF- Abandon Fields, HPF- Human Planted Forest, HS- Human Settlements,
DsS- Dwarf-shrub Steppe and NF- Natural Forest.
I also compared the change in the direction of correlation of each specific variable across observation
radii. I found (Table 3.1.2.2) that the change in the correlation direction differed between the examined
variables. The correlation of PA, HPF, HS and NF with mammal road-kills was consistent across
different radii (positive for PA, HPF and HS and negative for NF). The correlation between AA and
mammal road-kill numbers was negative at two observation radii (100 and 2000 m) and positive at all the
rest. As AA exhibited the highest importance index value at an observation radius of 2000 m, it is
suggested that the negative correlation at this radius had a stronger effect on road-kills of mammals than
its positive correlation with road-kills found at other radii. Lastly, the correlation directions of AF and
36
DsS with mammal road-kills changed as observation radii increased (from negative to positive for AF and
vice versa for DsS). Both land-uses -- AF and DsS -- exhibited highest importance index values at
observation radii at which they exhibited the negative correlation (i.e. 100 and 2000 m for AF and DsS,
respectively). As such, my results imply that for AF and DsS the negative correlation at these observation
radii has a more profound effect on road-kills of mammals than the positive correlation observed at other
radii.
Thereafter, I selected each variable at the radius at which it exhibited the highest importance value. I
then constructed a new set of models composed of all the combinations of these variables (with no
interactions). Table 3.1.2.3 presents variables' importance index values and estimated coefficients.
Table 3.1.2.3: Summary of the importance index values and coefficient estimates of the variables'
effect on mammal road-kill numbers, for the landscape scale. Variables included in the models are at
the radius in which they previously showed highest importance index values. PA- Perennial
Agriculture, AA- Annual Agriculture, AF- Abandon Fields, HPF- Human Planted Forest, HS- Human
Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Class
Variable
Mammals
Radius of observation
Importance
Coef_est
AF
100
0.99
-2.44
NF
2000
0.99
-10.45
DsS
2000
0.99
-5.27
AA
2000
0.72
-1.68
PA
500
0.53
0.86
HS
500
0.40
0.57
HPF
500
0.28
0.20
The most dominant observation radii were 2000 m and 500 m, with each being represented by 3
variables out of 7. The most important landscape-scale variables for predicting road-kill numbers of
mammals were AF, NF and DsS (importance index value of 0.99 for all). I also found a moderate
importance (0.72) for cover percentage of AA. Furthermore, all four variables exhibited a negative
correlation with mammal road-kill numbers. This means that lower cover percentage of these land-uses
around road-sections results in higher numbers of mammal road-kills.
Multi-scale effects
Based on the scale-dependent approach I aimed to achieve a more comprehensive understanding of
the overall effects of different variables from the two examined scales on mammal road-kill numbers.
Thus, I examined the relative joint contribution of local- and landscape-scale variables, i.e. multi-scale
37
effects. I generated and analyzed 128 models composed of all possible combinations (with no
interactions) of seven variables from different scales. I included three variables from the local scale and
four from the landscape scale. I selected the variables based on the previous analyses conducted for each
scale separately. These variables exhibited the highest importance index values in their related scale (see
above). The local scale variables I included were illumination, road-surface conjunction and distance to
the nearest passage. The landscape variables that I included were land-use covers of DsS, NF and AA, at
observation radius of 2000 m and AF at observation radius of 100 m.
Results resembled those obtained for bird road-kill numbers. Overall, landscape scale variables had
higher importance values in determining numbers of mammal road-kills (Table 3.1.2.4). Also, as for
birds, mammal road-kills were also profoundly affected by the presence of illumination. However, unlike
in birds, the road-surface conjunction was also of high importance (0.93) in influencing numbers of
mammal road-kills.
Table 3.1.2.4: Summary of the multi-scale analysis for mammal road-kills. Table shows the relative
importance and correlation trend of the examined variables. AF- Abandon Fields, DsS- Dwarf-shrub
Steppe, NF- Natural Forest and AA- Annual Agriculture.
Variable
Illumination
Road-surface conjunction
Distance to nearest passage
AF
DsS
NF
AA
Scale
Local
Local
Local
Landscape
Landscape
Landscape
Landscape
Radius of
observation (m)
NA
NA
NA
100
2000
2000
2000
38
Importance
Correlation
0.99
0.93
0.34
0.99
0.99
0.98
0.98
+
-
3.1.3 Reptiles
Spatial distribution
According to ANNI analysis, road-kill patterns of reptiles were aggregated, for all 3 examined roads
(Table 3.1.3.1). The ANNI for all the examined roads differed significantly from the expected value for
random or over-dispersed patterns (N=60, N=61, N=46, for roads 3, 44 and 424, respectively; P<0.0001
for all roads). Thus, I can conclude that road-kills of reptiles are also spatially clustered in all the
examined roads.
Table 3.1.3.1: Summary of the average nearest neighbor analysis for road-kills pattern of reptiles.
Reptiles
Observed Mean Distance
Expected Mean Distance
Nearest Neighbor Ratio
z-score
p-value
Road 3
136.523651
216.839022
0.629608
-5.488678
<0.0001
Road 44
52.770428
139.230401
0.379015
-9.278480
<0.0001
Road 424
97.903537
183.372018
0.533907
-6.047595
<0.0001
As before, the next spatial analysis I conducted was Ripley's K-statistics. Similar to the results for
birds' and mammals' road-kills, I found that road-kills of reptiles were spatially aggregated in all three
roads. Likewise, the spatial range at which road-kill aggregations of reptiles differed between the
examined roads (Figure 3.1.3.1). Road-kills were spatially aggregated up to an observation range of ~400
m in roads ‘44’ and ‘424’, in comparison with an observation range of up to 750 m in road ‘3’. This
implies that road ‘3’ imposes a more profound fragmentation on the landscape, funneling movement of
reptiles as well, to more specific locations.
39
Road 44
Road 424
~ 450 m
~ 550 m
Road 3
~ 1000 m
------ Expected K
------ Observed K
------ Confidence Env.
Figure 3.1.3.1: Spatial analysis of reptile road-kills aggregations utilizing a common
transformation of Ripley's k-function where the expected result with a random set of points is
equal to the input distance.
Same as before, I further examined the effect of the landscape heterogeneity on these aggregation
patterns based on the clustering results. In accordance with the scale-dependence approach, the following
sub-sections describe my analyses in two main spatial scales -- local and landscape -- and in a multi-scale
manner (i.e. including variables from both scales).
40
Local scale effects
I did not find a significant effect of illumination on reptile road-kill numbers (p=0.27; Figure 3.1.3.2).
However, I did find an effect of the relative level of the road compared to its surroundings (Figure
3.1.3.3). Consistent with the results for bird and mammal road-kills, significantly higher numbers of
reptile road-kills were found in road-sections which were at the same level as their surroundings, in
Number of road-kills
comparison with unleveled sections (p<<0.01).
p = 0.27
Illumination:
Absence = 0 Presence = 1
Number of road-kills
Figure 3.1.3.2: The effect of illumination on reptile road-kill numbers.
p<<0.01
Road-surface conjunction:
Even = 0
Uneven = 1
Figure 3.1.3.3: The effect the relative level of the road with respect to its surroundings (e.g. roadsurface conjunction) on reptile road-kill numbers.
The distance to the nearest road-crossing passage was significantly correlated with reptile road-kills.
This correlation presented a logarithmic decline (p<<0.001; Figure 3.1.3.4). This means that reptile roadkills were also, as for birds and mammals, found more close-by to passages rather than far away from
them.
41
90
80
Number of road0kills
70
60
50
40
y = -22.56ln(x) + 54.294
R² = 0.7644
p<<0.001
30
20
10
0
Distance to nearest crossing passage (m)
Figure 3.1.3.4: The effect of the distance to the nearest road-crossing passage on reptile road kill
numbers.
Landscape scale effects
Table 3.1.3.2 shows a summary of the importance index values of the examined landscape scale
variables at each observation radius, for reptile road-kills. My analysis revealed that land-use cover of PA
was of high importance at all of the observation radii. I also found that AA was of high importance at all
of the observation radii but 2000 m. Furthermore, DsS exhibited relatively high importance index values
at observation radii of 100 and 200 m. AF was of high importance at 500 m. At 1000 m, all variables
were of relative high importance, suggesting a mixed effect of the combination of several land-use covers
at this observation radius. NF exhibited high importance in predicting reptile road kill numbers at an
observation radius of 2000 m.
42
Table 3.1.3.2: Summary of the sets of models for the 5 landscape observation radii. Table shows
importance index values of the landscape scale variables on road-kill numbers and the AICc weight of
the best model in each set. The importance index values of variables which were included in the best
model are marked with *. The trend of the variables' estimated coefficients are color coded with red
and blue for positive and negative correlation, respectively. The examined landscape variables are:
PA- Perennial Agriculture, AA- Annual Agriculture, AF- Abandon Fields, HPF- Human Planted
Forest, HS- Human Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Class
Reptiles
Buffer_r
PA
AA
100
0.95*
0.72*
250
*
*
*
*
0.97
*
0.80
AF
HPF
HS
DsS
NF
Best model AICc wt.
0.51*
0.45*
0.48*
0.74*
0.64*
0.18
0.46
*
0.54
*
0.78
0.45
0.13
0.52
500
1.00
0.75
0.74
0.39
0.41
0.39
0.49
0.11
1000
0.96*
0.72*
0.63
0.72*
0.70*
0.64
0.56
0.13
2000
0.79*
0.38
0.34
0.33
0.33
0.42
0.97*
0.12
Comparing the change in variables’ importance index values across observation radii for each
examined land-use cover (Figure 3.1.3.5), I found different trends for the different relative land-use
covers. The importance index values of PA exhibited a concave (n-shape) correlation with radius of
observation (p=0.007, R2=0.99). The AA land-use cover exhibited a negative correlation with
observation radius (p=0.03, R2=0.83).
Additionally, I found that the change in the correlation's direction differed between the examined
variables (Table 3.1.3.2). The PA, AA and HPF correlations were consistently positive. The direction of
correlation between reptile road-kills and the land-uses AF, HS, DsS and NF changed as observation radii
increased. However while the correlation direction for AF changed from positive to negative at 250 m,
for DsS this direction change occurs at 2000 m. As such, most coefficients for AF were negative while
for DsS they were positive, including the radius at which each land-use exhibited highest importance (500
and 250 m, for AF and DsS, respectively). Thus, results suggest that the negative correlation for AF and
the positive correlation for DsS are of higher impact on road-kills of reptiles than correlations found at
other radii.
Similarly, HS exhibited mostly positive coefficients, including the positive coefficient at 1000 m (the
radius of observation at which HS has highest importance index). This implies that the positive
correlation between HS and road-kills of reptiles is of higher impact than the negative correlation found at
observation radius of 250 m. In a similar manner, NF exhibited negative coefficients at all observation
radii but 100 m, including the radius at which it exhibited the highest importance index (2000 m). This
suggests that the negative correlation between NF and road-kills of reptiles was more substantial than the
positive one.
43
PA
AF
1
1
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
y = -1E-07x2 + 0.0001x + 0.94
R² = 0.99
p = 0.007
0.5
0.4
0.3
Importance index value
AA
1
0.6
0.5
0.5
y = -0.0002x + 0.83
R² = 0.83
p = 0.03
0.4
0.3
0.4
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
0
500
1000
1500
2000
0
0
500
1000
HPF
1500
2000
0
1
1
1
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
0.5
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
500
1000
1500
2000
1500
2000
1000
1500
2000
1500
2000
NF
0.9
0
500
DsS
0
0
500
1000
1500
2000
0
500
1000
HS
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
500
1000
Radius of observation (m)
Figure 3.1.3.5: The effect of landscape scale variables (cover percentage of different land-uses) on
reptile road-kills. Graphs show importance index values as a function of observation radius, for each
examined variable. The examined landscape variables are: PA- Perennial Agriculture, AA- Annual
Agriculture, AF- Abandon Fields, HPF- Human Planted Forest, HS- Human Settlements, DsSDwarf-shrub Steppe and NF- Natural Forest.
I selected each variable at the radius at which it exhibited the highest importance index value and
designed a new set of models based on all the combinations of these variables (with no interactions). This
analysis showed (Table 3.1.3.3) that the most important landscape-scale variables for predicting reptile
road kill numbers were NF and PA (importance index value of 1 for both). Moreover, AA and DsS also
showed high importance in predicting reptile road kill numbers (index values of 0.9 and 0.82,
respectively). Analysis of the estimated coefficients revealed a negative correlation between reptile road
kill numbers and NF cover, and in contrast, positive correlations with PA, AA and DsS covers. This
means that lower cover percentage of NF around road-sections makes them more prone to road-kills of
44
reptiles. Conversely, as the relative cover of PA and/or AA and/or DsS increased so did reptile road kill
numbers.
Table 3.1.3.3: Summary of the importance indices and coefficients estimates of landscape scale
variables' effect on reptile road kill numbers. Each variable included in the models was within the
radius at which it had the highest importance index. PA- Perennial Agriculture, AA- Annual
Agriculture, AF- Abandon Fields, HPF- Human Planted Forest, HS- Human Settlements, DsSDwarf-shrub Steppe and NF- Natural Forest.
Class
Variable
Radius of observation
Importance
Coef_est
NF
2000
1.00
-16.84
PA
500
1.00
3.21
AA
250
0.90
1.25
DsS
250
0.82
1.91
HPF
1000
0.46
1.54
AF
500
0.45
-1.62
HS
1000
0.30
-0.38
Reptiles
Multi-scale effects
With respect to the scale-dependent approach, I aimed for a better understanding of the overall effects
of different factors from different scales on reptile road-kill numbers. Thus, I analyzed the relative joint
contribution of local- and landscape-scale variables, i.e. multi-scale effects. I generated and analyzed 128
models composed of all possible combinations (with no interactions) of seven variables from different
scales. I included the three variables from the local scale and four variables from the landscape scale.
Again, I selected the variables based on the analyses I previously conducted for each scale separately.
These variables exhibit the highest importance index values in their related scale (see above). The local
scale variables I included are illumination, road-surface conjunction and distance to the nearest passage.
The landscape variables which I included are land-use covers of NF at observation radius of 2000 m, PA
at observation radius of 500 m and AA and DsS at observation radius of 250 m.
Results showed (Table 3.1.3.4) high relative importance for variables from both of the spatial scales.
Beside 2 variables (AA and DsS), all of the examined variables had relatively high importance. First, at
the local scale, road-surface conjunction and distance to the nearest crossing passage had the highest
importance in predicting reptile road-kill numbers (1 for both). Landscape scale variables exhibited the
next highest importance in determining rates of reptile road-kills (0.98 and 0.94 for NF and PA,
respectively). Lastly, reptile road-kill numbers were also profoundly affected by the presence of
45
illumination (importance index value = 0.86). This means that road-kills of reptiles are effected by both
local- and landscape-scale variables. Thus, the resulted pattern of reptile road-kills is determined by a
multi-scale effect.
Table 3.1.3.4: Summary of the multi-scale analysis for reptile road-kill numbers. Table shows the
relative importance and correlation trend of the examined variables. NF- Natural Forest, PAPerennial Agriculture, AA- Annual Agriculture and DsS- Dwarf-shrub Steppe.
Variable
Road-surface conjunction
Distance to nearest passage
Illumination
NF
PA
AA
DsS
Scale
Local
Local
Local
Landscape
Landscape
Landscape
Landscape
Radius of
observation (m)
NA
NA
NA
2000
500
250
250
46
Importance
Correlation
1
1
0.86
0.98
0.94
0.42
0.31
+
+
+
+
3.2 Passage-use rate patterns
My second study question refers to the determinants affecting passage use rates. First I examined the
time preferences of species using the road-crossing passages. Based on previous studies (e.g. Little et al.
2002) I separated between the major trophic associations, i.e., between predators and prey. I obtained
totals of 90 and 52 observations of predators and prey species crossing at the passages, respectively.
Predator species mainly included Egyptian mongoose (Herpestes ichneumon; 64 observations) and golden
jackal (Canis aureus; 18 observations). Other predators observed were red fox (Vulpes vulpes), domestic
cat (Felis silvestris catus) and even one observation of a striped hyena (Hyaena hyaena). Prey species
mainly included black rat (Rattus rattus; 20 observations) and Indian porcupine (Hystrix indica; 16
observations). Other prey species observed were brown rat (Rattus norvegicus), house mouse (Mus
musculus), roughtail rock agama (Laudakia stellio) and even two observations of snakes (unidentified).
Representative photo captures are presented in Figure 3.2.1.
Laudakia stellio
Herpestes ichneumon
Herpestes ichneumon
Hystrix indica
Rattus rattus
Canis aureus
Hyaena hyaena
Figure 3.2.1: Selected photo captures of species using the examined road-crossing passages.
I found that predators preferred daytime (54 observations at daytime compared to 36 observations at
night; Chi square, p=0.057) whereas prey species preferred nighttime (44 observations at daytime
compared to 8 observations at night; Chi square, p<0.001; Figure 3.2.2).
47
Figure 3.2.2: Number of crosses at road-passages by predator and prey species, at daytime and at
nighttime
In order to examine the effect of spatial heterogeneity on wildlife rates of use of these passages I
generated and analyzed a total of 864 General Linear Models (GLMs). I included in this analysis only
passages with at least 6 inspection periods (see chapter 2). In the next sections I present my results in
accordance with the spatial-oriented approach used in this study. The first section shows my analysis of
the local attributes and their effect on wildlife passage-use rates. The second section presents the effect of
landscape scale attributes at 5 different radii of observation. Finally, the third section presents a multiscale analysis I conducted in order to identify scale-dependent patterns.
Local scale effects
In order to identify the variables explaining wildlife passage-use rates at the local scale I applied the
model selection procedure. I constructed and analyzed two sets of models for passage-crossing
observations of: a) predators species; b) prey species. Each set contained 16 models, constructed from all
the combinations of 4 local scale variables – 'Plant cover' (%), 'Openness', 'Human use level' (categorical
variable which reflects the level of usage in these passages by humans) and 'Distance to closest
neighboring passage' (m). I determined the most relevant models by their AICc values. Table 3.2.1
presents the relevant top models (e.g. models with delta AICc < 3; Anderson and Burnham 2002) for each
set.
48
Table 3.2.1: The effect of local scale variables on passage use rates. Table shows best models based
on AICc values (e.g. models with delta AICc < 3; Anderson and Burnham 2002). Variables included
or excluded in the model are indicated with 1 or 0, respectively. Variables abbreviations: HU =
Human use, OP = Openness, PC = Plant cover and CNP = Closest neighboring passage.
Species
Model_name K AICc
Predators M_Local_10 6 -5.26
M_Local_13 7 -4.48
M_Local_14 7 -3.80
Prey
M_Local_16 2 2.69
3 3.58
M_Local_1
4 4.37
M_Local_5
3 5.17
M_Local_3
3 5.41
M_Local_2
∆AICc
0.00
0.78
1.45
0.00
0.89
1.67
2.47
2.72
AICcWt HU OP PC CNP
0.43
1
1
0
0
0.29
1
1
1
0
0.21
1
1
0
1
0.34
0
0
0
0
0.22
0
0
1
0
0.15
0
0
1
1
0.10
0
1
0
0
0.09
0
0
0
1
I further calculated the importance index of each variable and examined its trend of effect (i.e.
direction of correlation; Table 3.2.2). For predatory species, 'Human use' and 'Openness' were the most
important variables influencing model performance (importance index values of 0.96 and 0.98,
respectively). These two variables were included in all of the 3 top best-performing models. For prey
species, I found that the best performing model was the null model (AICc weight = 0.34). Nevertheless,
from comparing the relative importance of the explanatory variables, I found that prey species’ passageuse was mostly affected by the plant cover at its vicinity (importance index value of 0.45).
Table 3.2.2: Effect of local scale variables on passage use rates of predators and prey species. The
importance index values and correlation trend of the examined local variables are shown.
Species
Predators
Prey
Variable
Human use level
Openness
Plant cover
Closest neighboring
passage
Human use level
Openness
Plant cover
Closest neighboring
passage
Importance index
0.96
0.98
0.35
0.24
Correlation
Negative
Positive
Positive
Positive
0.01
0.2
0.45
0.28
Negative
Positive
Positive
Negative
Exploring the coefficient of each variable (table 3.2.2) revealed that openness and plant cover were
positively correlated with passage-use rates of all examined species. In contrast, human use of the
passages was negatively correlated with passage-use rates of the surveyed species. I also found that the
49
distance to the closest nearest neighboring passage was negatively correlated with passage-use rates of
prey species, but positively correlated with use rates of predatory species. Thus, my analysis suggests
that shorter distances to the nearest neighboring passage result in higher rates of use by prey species and
lower rates of use by predators.
Landscape scale effects
In order to examine the effect of landscape scale variables on passage-use rates of predators and prey
species I conducted the following analyses. First, I calculated the cover percentage of different land-uses
at increasing radii of observation (e.g. 100, 250, 500, 1000 and 2000 meters) around passages. Then, I
generated and analyzed 5 sets of models (i.e. a set for each observation radius). Each set contained 64
models based on the 6 following landscape land-use variables: perennial agriculture (PA), annual
agriculture (AA), human planted forest (HPF), human settlement (HS), dwarf-shrub steppe (DsS) and
natural forest (NF). I further examined the importance index values of these variables at each observation
radius and for each examined species group (e.g. predators and prey; Table 3.2.3).
Table 3.2.3: Summary of the sets of models for the 5 landscape observation radii. Table shows
importance index values of the landscape scale variables’ effect on passage-use rates and the AICc
weight of the best model in each set. Importance indices of variables which are included in the best
model are marked with *. The trend of the variables' respected coefficients are color coded with red
and blue for positive and negative correlation, respectively. The examined landscape variables are:
PA- Perennial Agriculture, AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human
Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Diet
Buffer_r
PA
AA
HPF
HS
DsS
NF
100
0.20
0.43
0.33
0.25
0.61*
0.50*
0.16
0.23
*
*
0.20
*
0.14
*
250
Predators
Prey
500
0.20
0.19
0.25
0.19
0.29
0.34
0.17
*
0.64
0.22
Best model AICc wt.
0.59
0.47
1000
0.20
0.20
0.21
0.40
0.17
0.59
0.14
2000
0.25
0.23
0.21
0.38
0.17
0.28
0.16
100
0.39
0.53*
0.18
0.45*
0.22
0.78*
0.11
250
0.32
0.65*
0.22
0.62*
0.16
0.75*
0.19
0.21
*
0.14
*
0.17
*
0.15
500
1000
2000
0.27
0.38
0.37
0.27
0.25
*
0.35
0.22
0.22
0.20
0.52
0.24
0.18
0.31
0.24
0.82
0.90
0.73
Land-use cover of NF was of high importance in determining models' performance for predicting
passage-use rates. This observation was consistent in all of the observation radii, both for predators and
prey species. DsS had a strong effect on passage-use rates of predators (importance index values of 0.61
and 0.64 at observation radii of 100 and 250 m, respectively), likewise for HS (importance index of 0.4 at
50
an observation radius of 1000 m). Passage-use rates of prey species were moderately affected by the
land-use covers of HS and AA (mainly at observation radius of 250 m with importance index values of
0.62 and 0.65, respectively).
I further compared the change in variables’ importance index values across observation radii for each
examined land-use cover (Figures 3.2.3 and 3.2.4). I conducted this comparison in order to identify the
radius of observation at which each specific variable had the most profound effect and to examine trends
in effect of specific variables as observation radius changed. The importance index of only one variable
(the effect of PA on passage-use rates of predators) was significantly correlated with radius of observation
(p = 0.004, R2 = 0.99). This correlation exhibited a polynomial convex (U-shape) link.
Importance index value
PA
AA
HPF
1
1
1
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
0.5
y = 3E-08x2 - 4E-05x + 0.21
R² = 0.99
p = 0.004
0.4
0.3
0
500
1000
1500
2000
0
0
500
HS
1000
1500
2000
0
DsS
1
1
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
0.5
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
500
1000
1500
2000
1000
1500
2000
NF
1
0
500
0
0
500
1000
1500
2000
0
500
1000
1500
2000
Radius of observation (m)
Figure 3.2.3: The effect of landscape scale variables (cover percentage of different land-uses) on
passage use rates of predators. Graphs show importance index values as a function of observation
radius, for each examined variable. The examined landscape variables are: PA- Perennial Agriculture,
AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human Settlements, DsS- Dwarf-shrub
Steppe and NF- Natural Forest.
51
PA
Importance index value
1
AA
1
0.9
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
0.5
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
0
500
1000
1500
2000
HS
1
0
0
500
1000
1500
2000
0
DsS
1
0.9
0.9
0.8
0.8
0.8
0.7
0.7
0.7
0.6
0.6
0.6
0.5
0.5
0.5
0.4
0.4
0.4
0.3
0.3
0.3
0.2
0.2
0.2
0.1
0.1
0.1
0
0
500
1000
1500
2000
500
1000
1500
2000
NF
1
0.9
0
HPF
1
0
0
500
1000
1500
2000
0
500
1000
1500
2000
Radius of observation (m)
Figure 3.2.4: The effect of landscape scale variables (cover percentage of different land-uses) on
passage use rates of prey species. Graphs show importance index values as a function of observation
radius, for each examined variable. The examined landscape variables are: PA- Perennial Agriculture,
AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human Settlements, DsS- Dwarf-shrub
Steppe and NF- Natural Forest.
I also examined correlations' directions based on their coefficients (positive coefficient for positive
correlation and vice versa). I compared the trend of the correlations' direction of effect of each specific
variable across observation radii (Table 3.2.3). The trend of the correlations' directions across
observation radii differed between the examined variables -- while PA, HPF and NF had a constant effect
on passage-use rates of predators (positive correlations for NF and negative for PA and HPF), the
correlations of AA, HS and DsS changed their direction of effect on passage-use rates of predators as
radius of observation increased. Consequently, I examined at which radius each variable was of highest
importance, as well as which correlation direction was found at more observation radii. I found that AA
exhibited a more profound effect of the negative correlation with passage-use rates of predators, whereas
HS and DsS showed greater effect of the positive correlation. This means that lower cover of AA around
passages resulted in higher rates of use by predators, while higher covers of HS and DsS around passages
resulted in higher rates of use by predators. For example, passages which were located near settlements
were more frequently use by predators. On the other hand, passages located near wheat fields (for
52
example), were less frequently used by predator species such as Egyptian mongoose (Herpestes
ichneumon) and golden jackal (Canis aureus).
For the prey species, correlation directions were consistent for AA, DsS and NF (positive correlations
for DsS and NF and negative for AA). Conversely, I found that the direction of correlations between
passage-use rates of prey and PA, HPF and HS changed as observation radius increased. As before, I
examined at which radius each variable was of its highest importance, as well as which correlation
direction was found at more observation radii. PA was more positively correlated with passage-use rates
of prey species, whereas HPF and HS were more negatively correlated. These results suggest that
passages located near human settlements and/or human planted forests are used less frequently by prey
species. Moreover, passages located in the vicinity of PA (such as Orange plantations) are more
frequently used by prey species such as black rat (Rattus rattus) and Indian porcupine (Hystrix indica).
Following these results, I selected each variable at the radius at which it exhibited the highest
importance value. I then constructed a new set of models using all the combinations of these variables
(with no interactions). Table 3.2.4 presents variables' importance index values and estimated coefficients.
Passage-use rates of predators were mainly influenced by four land-uses: AA and HPF (importance
index values of 0.54 and 0.53, respectively) were negatively correlated with passage-use rates of
predators, whereas DsS and NF (importance index values of 0.47 and 0.43, respectively) were positively
correlated with these rates. For prey species, passage-use rates were positively correlated with NF
relative cover (importance index of 0.57) and negatively correlated with AA and HS covers (importance
index values of 0.64 and 0.63, respectively). This means that lower cover percentage of AA around
crossing passages increased their rates of use (both by predators and by prey species). Furthermore, lower
cover percentages of HPF (for predators) and HS (for prey) also increased passage rates of use. On the
other hand, passage-use rates were positively correlated with the relative covers of DsS (for predators),
AA (for prey) and NF (for both). Therefore, as the relative cover of these land-uses increased, so did
passage-use rates of the related species.
53
Table 3.2.4: Summary of the importance index values and coefficient estimates of the effect of
landscape scale variables on passage-use rates. Each variable included in the models was within the
radius at which it had the highest importance index. The examined landscape variables are: PAPerennial Agriculture, AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human
Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Diet
Predators
Prey
Variable
Radius of observation
Importance
Coef_est
AA
100
0.54
-0.44
HPF
500
0.53
-0.83
DsS
250
0.47
1.17
NF
250
0.43
1.19
PA
2000
0.29
-0.67
HS
1000
0.21
0.67
NF
1000
0.77
4.50
AA
250
0.64
-0.47
HS
250
0.63
-1.03
PA
100
0.33
0.32
DsS
1000
0.26
0.98
HPF
500
0.18
-0.17
Multi-scale effects
In accordance with the scale-dependent approach of this study in general, and with the multi-scale
hypothesis specifically, I explored the overall effects of different variables from the two examined scales
on passage-use rates. More specifically, I looked at the relative joint contribution of local- and landscapescale variables. Therefore, I generated and analyzed two sets of models. The first set, for predator
species, included 128 models composed of all possible combinations (with no interactions) of 7 variables
from different scales (3 local and 4 landscape scale). The local scale variables included in this model
were 'Human use level', 'Openness' and 'Plant cover'. The landscape variables which I included were
land-use covers of AA, DsS, NF and HPF at observation radii of 100, 250, 250 and 500 m, respectively
(Table 3.2.5).
The second set, for prey species, included 64 models composed of all possible combinations (with no
interactions) of 6 variables from different scales. I included 3 variables from each scale (e.g. local and
landscape). The local scale variables were 'Openness', 'Plant cover' and 'distance to the nearest passage'.
The landscape variables were land-use covers of AA at observation radius of 250 m, NF at observation
radius of 1000 m and HS at observation radius of 250 m (Table 3.2.5).
54
Table 3.2.5: The effect of spatial context on passage-use rates, summary of the multi-scale analysis.
Table shows the relative importance and correlation trend of the examined variables. The examined
landscape variables are: AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human
Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Diet
Variable
Predators Human use level
Openness
Plant cover
AA
DsS
NF
HPF
Prey
Openness
Plant cover
Closest neighboring passage
AA
HS
NF
Scale
Local
Local
Local
Landscape
Landscape
Landscape
Landscape
Local
Local
Local
Landscape
Landscape
Landscape
Radius of
observation (m)
NA
NA
NA
100
250
250
500
NA
NA
NA
250
250
1000
Importance
Correlation
0.84
0.71
0.36
0.27
0.35
0.19
0.35
0.14
0.35
0.49
0.63
0.52
0.82
+
+
+
+
+
+
+
Passage-use rates of predators were more influenced by local scale variables, while passage-use rates
of prey species were more affected by landscape scale variables.
55
3.3. Predation pressure patterns
My third study question refers to the determinants affecting predation pressure in the vicinity of a
passage. In order to quantify predation pressure, I used a known practice of Artificial Nest Predation
(ANP; Soderstrom et al. 1998). This ANP procedure relies on documenting predation events. Thus, I
documented presence or absence of placed eggs (i.e. predation events) in a binomial manner. The
frequency of absent eggs (hereafter, predation occurrence) reflected the predation pressure. Accordingly,
based on the collected data of predation occurrence in the vicinity of 10 potential road-crossing passages,
I generated and analyzed 576 General Linear Models (GLMs).
With respect to my null hypothesis (the random pattern hypothesis), I generated and analyzed a
null model. This model examines whether passages significantly differ from one another in predation
occurrence and if so, can this pattern be explained by a random effect. The statistical expression of this
model was: Eggs_eaten ~ (1|Passage_index); whereas Eggs_eaten is the binary variable which represents
predation occurrence and Passage_index stands for a categorical index to each examined passage. I found
that passages significantly differ one from another in terms of predation occurrence and frequency
(p<0.01). Furthermore, the observed pattern of predation occurrence at the examined passages
significantly differ from an expected pattern resulted from a random effect (p<<0.01).
Thereafter, I used sets of GLMs to identify spatial attributes that influence predation occurrence. In
the next sections I present my results in accordance with the spatial scale. The first section shows the
analysis of the local attributes and their effect on predation pressure at the vicinity of road-crossing
passages. The second section presents the results for the effect of landscape scale attributes at five
different radii of observation. The third section presents a multi-scale analysis I conducted in order to
identify scale-dependent patterns. The fourth and last section of the results exhibits findings regarding the
effect of the distance between neighboring passages (i.e. passages density) on the predation pressure.
Local scale
I constructed and analyzed 64 models from 6 local scale variables – 'Plant cover' (%), 'Open length'
(m), 'Height from the road surface' (m), 'Openness', 'Human use level' (categorical variable which reflects
the level of usage in these passages by humans) and 'Distance to closest neighboring passage' (m). I
determined the most relevant models by their AICc values. Table 3.3.1 presents the relevant top models
(e.g. models with delta AICc < 3; Anderson and Burnham 2002).
56
Table 3.3.1: The effect of local scale variables on predation occurrence at passages. Table shows
best models based on AICc values (e.g. models with delta AICc < 3; Anderson and Burnham 2002).
Variables included or excluded in the model are indicated by 1 or 0, respectively. Variable
abbreviations: HU = Human use level, PC = Plant cover, HFR = Height from road, OP = Openness,
CNP = Closest neighboring passage, OL = Open length.
Model name
K
AICc
∆AICc
AICcWt
HU
PC
HFR
OP
CNP
OL
M_local_30
6
144.62
0.00
0.17
1
1
1
0
0
0
M_local_40
6
145.87
1.25
0.09
1
0
0
1
0
1
M_local_31
6
146.30
1.68
0.07
1
1
0
1
0
0
M_local_39
6
146.63
2.01
0.06
1
0
1
0
0
1
M_local_46
7
146.74
2.12
0.06
1
1
1
0
1
0
M_local_51
7
146.77
2.15
0.06
1
1
1
1
0
0
M_local_49
7
146.79
2.17
0.06
1
1
1
0
0
1
M_local_18
5
147.35
2.73
0.04
1
0
0
0
0
1
In these models, three variables were consistently included: plant cover percentage, human use level,
and the height from road surface. In turn, I found (Table 3.3.2) that these three variables had the highest
importance index values for the model performance (0.97, 0.66 and 0.58 for human use level, plant cover
and height from road, respectively). This observation is consistent with the results from my AICc
analysis, which showed that the best model (e.g. lowest AICc; M_local_30) was composed from these
same three variables.
Table 3.3.2: The effect of local scale variables on predation occurrence at passages. Table shows the
importance index and correlation direction of the examined local variables.
Variable
Human use level
Plant cover
Height from road
Open length
Openness
Closest neighboring passage
Importance index
0.97
0.66
0.58
0.52
0.46
0.30
Correlation
+
+
I also examined the direction of correlation of each variable. I found that plant cover was positively
correlated with predation occurrence. On the other hand, human use level and height from the road
surface were both negatively correlated with predation occurrence. I found the effect of human use level,
which was a categorical variable, to be statistically significant (p<0.01) only for the highest level of use
(e.g. at least once a day).
According to these results, the strong negative correlation of human use with predation occurrence
suggests that higher usage of passages by humans leads to lower rates of use by predator species and in
57
turn, lower rates of predation. Height from the road surface is also negatively correlated with predation
pressure, implying that when movement is more difficult, predation pressure decreases. Lastly, I found
that plant cover was positively correlated with predation occurrence, indicating that high and dense
vegetation around the passages encouraged predators to increase their activity at these locations.
Landscape scale effects
In order to examine the effect of landscape scale variables on predation occurrence I conducted the
following analyses. First, I calculated the cover percentage of different land-uses at increasing radii of
observation (i.e. 100, 250, 500, 1000 and 2000 meters) around passages. Then, I generated and analyzed
five sets of models (i.e. a set for each observation radius). Each set contained 64 models based on the 6
following landscape land-use variables: perennial agriculture (PA), annual agriculture (AA), human
planted forest (HPF), human settlement (HS), dwarf-shrub steppe (DsS) and natural forest (NF). I further
examined the importance index values of these variables at each observation radius (Table 3.3.3).
Table 3.3.3: Summary of the sets of models for the five landscape observation radii. Table shows
importance index values of the landscape scale variables effect on predation occurrence and the AICc
weight of the best model in each set. The importance indices of variables which are included in the
best model are marked with *. The examined landscape variables are: PA- Perennial Agriculture,
AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human Settlements, DsS- Dwarf-shrub
Steppe and NF- Natural Forest.
Buffer_r
PA
AA
HPF
HS
DsS
NF
Best model AICc wt.
100
0.39
0.33
0.52*
0.98*
0.86*
0.34
0.15
250
0.80*
0.64*
0.36
0.85*
0.74
0.62*
0.17
500
0.39
0.35
0.75*
0.29
0.36
0.34
0.10
1000
0.58*
0.57*
0.97*
0.53*
0.63*
0.62*
0.32
2000
0.96*
0.88*
0.47
0.86*
0.27
0.75*
0.36
At small radii of observation within the landscape scale (100 and 250 m), predation occurrence was
best explained by percentage of HS cover (importance index values of 0.98 and 0.85, respectively). At an
observation radius of 100 m there was also high importance (0.86) of DsS cover, whereas at 250 m radius
there is high importance of PA cover (0.8). More variables which have moderate importance at an
observation radius of 250 m are DsS, AA and NF covers (importance index values of 0.74, 0.64 and 0.62,
respectively). At intermediate observation radii (500 and 1000 m), predation occurrence was best
explained by the relative cover of HPF (importance index values of 0.75 and 0.97, respectively). Lastly,
at the largest observation radius (2000 m) predation occurrence was best explained only when several
58
land-uses were taken into account. The most profound effects in this observation radius (2000 m) were
land-use covers of PA, AA and HS (importance index values of 0.96, 0.88 and 0.86, respectively).
I further compared the change in importance index values across observation radii for each examined
land-use cover (Figure 3.3.1). I conducted this comparison in order to identify the radius of observation
at which each specific variable had the most profound effect and to examine trends in effect of specific
variables as observation radius changed. No significant correlations were found between importance
index values and radii of observation.
AA
HPF
1
0.9
0.8
0.7
0.6
0.5
Importance index value
HS
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.4
0.3
0.2
0.1
0
0
500
1000
1500
2000
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
500
1000
DsS
1500
2000
0.8
0.9
0.7
0.8
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.1
0.1
0
0
1000
1500
2000
1500
2000
1500
2000
PA
0.2
0.2
1000
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.6
0.7
500
500
NF
1
0
0
0
500
1000
1500
2000
0
500
1000
Radius of observation (m)
Figure 3.3.1: The effect of landscape scale variables (cover percentage of different land-uses) on
predation occurrence at passages. Graphs show importance index values as a function of observation
radius, for each examined variable. The examined landscape variables are: PA- Perennial Agriculture,
AA- Annual Agriculture, HPF- Human Planted Forest, HS- Human Settlements, DsS- Dwarf-shrub
Steppe and NF- Natural Forest.
I next selected each variable at the radius at which it exhibited the highest importance value, and
constructed a new set of models built of all the combinations of these variables (with no interactions).
Next, I examined each variable's importance index and trend of effect (i.e. correlation direction) (Table
3.3.4). I found that predation occurrence in the vicinity of crossing passages is mainly influenced by four
land-uses. HS and HPF covers had the most profound effect (importance index values of 0.96 and 0.89,
respectively), with a positive correlation with predation occurrence for HS and a negative correlation for
59
HPF. There was a moderate influence for the land-use covers of PA and AA (both with an importance
index value of 0.61), both with a positive correlation with predation occurrence in the vicinity of crossing
passages. So, higher relative covers of HS, PA and AA resulted in higher predation occurrences.
Conversely, higher relative cover of HPF resulted in lower observed rates of predation.
Table 3.3.4: Summary of the importance index values and direction of correlation of landscape-scale
variables' effect on predation occurrence in the vicinity of crossing passages. Each variable included
in the models is within the radius at which it showed the highest importance index value. The
examined landscape variables are: PA- Perennial Agriculture, AA- Annual Agriculture, HPF- Human
Planted Forest, HS- Human Settlements, DsS- Dwarf-shrub Steppe and NF- Natural Forest.
Variable
Radius of observation (m)
Importance
Correlation
HS
100
0.96
+
HPF
1000
0.89
-
PA
2000
0.61
+
AA
2000
0.61
+
DsS
100
0.43
-
NF
2000
0.28
-
Multi-scale effects
I explored the overall effects of different variables from the two examined scales on predation
occurrence. More specifically, I looked at the relative joint contribution of local- and landscape-scale
variables. Therefore, in order to examine potential scale-dependent processes, I generated and analyzed
128 models composed of 7 variables from different scales (based on all possible combinations without
interactions). Table 3.3.5 presents the selected variables and their related importance index values and
trend of effect (direction of correlation).
I found that landscape scale variables had higher relative importance indices than local scale variables
(Table 3.3.5). Moreover, I found that at the landscape scale, the percentage of HS and agriculture cover
(both annual and perennial) were positively correlated with predation occurrence. In contrast, HPF is
negatively correlated with predation occurrence. At a local scale, plant cover was positively correlated
with predation occurrence, whereas height from the road surface and human use levels exhibited a
negative correlation with predation occurrence.
60
Table 3.3.5: The effect of spatial context on predation occurrence, summary of the multi-scale
analysis. Table shows the relative importance and correlation trend of the examined variables. The
examined landscape variables are: PA- Perennial Agriculture, AA- Annual Agriculture, HPF- Human
Planted Forest and HS- Human Settlements.
Variable
Scale
Plant cover
Height from road
Human use level
HS
HPF
AA
PA
Local
Local
Local
Landscape
Landscape
Landscape
Landscape
Radius of
observation (m)
NA
NA
NA
100
1000
2000
2000
Importance
Correlation
0.50
0.50
0.43
0.79
0.64
0.52
0.47
+
+
+
+
Proximity to neighboring passages
Although distance to neighboring passage did not have a significant effect on predation occurrence at
a local scale analysis, I conducted a different analysis to examine the effect of this variable on a larger
scale of observation. I conducted this additional analysis as this effect of neighboring passage might not
be related to a specific scale, but rather to the spatial context of the entire set of the passages. Thus, I
examined the effect of the density of passages on the change in predation occurrence at the vicinity of
road-crossing passages.
First, I calculated the predation frequency at each surveyed location, as the proportion of predation
occurrences of all visits (for example, five predation events out of 14 total visits = 0.36). Secondly, I
calculated the difference between predation frequencies (hereafter, Δf) at the most distant points (i.e., 200
m) and at the passage location itself (i.e., 0 m). The Δf could be either positive or negative, indicating
that predation frequency at distant locations was higher or lower from that at the structure itself,
respectively. I then examined the correlation between the change in predation frequency (i.e., Δf) and
proximity to the neighboring passage.
I found that Δf decreased significantly (p<0.01) with distance to the neighboring passage (Figure
3.3.2). Furthermore, the regression line, representing this decrease, crossed the X axis at around 800 m,
indicating a shift in the effect of the neighboring passage on predation pressure from positive (lower
pressure at the structure itself) to negative (higher pressure at the structure itself). This means that
predation frequency at distant locations shifts from higher values than predation frequency at the passage
itself, to lower values in comparison, as distance to neighboring passage increases. More specifically,
predation pressure was higher at the passages than far from them when passages were distant from each
61
other. On the other hand, predation pressure was higher far from the passages than adjacent to the
passages, when passages run close-by to each other.
R2 = 0.32
p < 0.01
y = -0.0003x + 0.25
Distance to neighboring passage (m)
Figure 3.3.2: Change in predation frequency (proportion of predation occurrence of all visits; Δf)
as a function of distance to the neighboring passage.
62
Chapter 4 – DISCUSSION
In this thesis, I measured and gathered data on various road permeability patterns and spatial variables
(factors), enabling me to evaluate the human-wildlife conflict that roads impose (i.e. the road-induced
fragmentation). As pointed out in the introduction chapter, a literature review reveals contradicting
results concerning the relative importance of local and landscape scale attributes affecting road
permeability patterns (e.g. road-kills as well as passage-use patterns). Whereas some studies have
claimed that local attributes are the most influential (Reed et al. 1975, Norman 1998, Cain et al. 2003),
other researchers have shown that the landscape matrix is the most important attribute (Foster and
Humphrey 1995, Yanes et al. 1995, Land 1996, Clevenger and Waltho 2000, Ng et al. 2004).
Accordingly, a front line debate in road ecology is whether local or landscape is the most important scale
for determining factors that influence permeability patterns. I advocate the possibility that different
focuses on spatial scales and levels of organization by different researchers produce different outcomes,
and consequently ambiguous conclusions.
As such, it should be beneficial to refer and study road permeability and its resulting patterns using a
scale-dependent approach. This approach represents the most up-to-date progress in spatial ecology and
is highly supported by cutting-edge developments of spatially-oriented technology, such as GIS
applications, remote sensing and geostatistical tools. Essentially, the scale-dependent approach
emphasizes that particular patterns and processes may depend on the spatial scale and biological
organizational level at which they are considered. For example, while a particular process (or
determinant) may well explain patterns of individual behavior at a given spatial scale, other processes (or
determinants) may explain population distribution at a larger spatial scale (Turner 2005b). In order for
this approach to be applied, a great deal of information is necessary. This includes, among others, data
regarding the organisms within the relevant area, road-kill locations, wildlife usage of potential roadcrossing passages and the spatial configuration of road and passage surroundings.
In line with applied ecology and conservation biology, although roads are a major and permanent
element in many landscapes, surprisingly few studies have specifically addressed the permeability of
roads and their role in habitat fragmentation (Alexander et al. 2005). As such, there is an urgent need to
expand our ecological knowledge, as well as to apply scientifically-based knowledge on road
permeability patterns. Moreover, applying this scientifically-based knowledge should benefit not only
biodiversity at different levels, but also save human life and injuries through minimizing wildlife-vehicle
collisions (Forman and Merrick 2003, Bellis et al. 2007).
This thesis examines the effect of roads on wildlife, through three complementary patterns, each
addressing a different aspect of road permeability and the effects of scale-dependent variables on these
63
patterns. The spatial scale over which I have been exploring these effects encompassed both local and
landscape scales. This multi-scale approach enabled me to examine how road permeability patterns are
affected by a long list of variables influencing several biological and/or ecological mechanisms such as
wildlife movement, wildlife decision-making, predator-prey interactions and biogeographical orientation.
Consequently, this approach allowed me to evaluate the effect of roads on wildlife by emphasizing the
prospective natural processes by which different species move in the environment and respond to the
existence of roads.
My thesis brings together scientific tools and knowledge of the spatial ecology approach, as well as
ideas and principles of nature conservation, in a human disturbed system (i.e. the road-landscape
network). This system brings into focus main issues of conservation ecology regarding human-wildlife
conflicts. More specifically, I have addressed road-induced fragmentation of the landscape and its
interference to wildlife movement, given its profound effects on wildlife persistence.
In this study I aimed to explore road permeability patterns. More specifically, I examined patterns of
road-kills and passage-use (both in means of rates and of predation pressure) in order to identify the
spatial attributes that influence these patterns. I integrated empirical and advanced theoretical methods to
study how local and landscape scale determinants affect road-kill and passage-use patterns of various
species in the Mediterranean landscape. My main objective was to identify spatial heterogeneity
attributes which determine:
a) Wildlife road-kill patterns;
b) Wildlife passage-use patterns;
c) Predation pressure patterns in the vicinity of road crossing passages.
There are many processes operating at different scales that may influence road-kills and usage of
passages by wildlife. Accordingly, I constructed a set of hypotheses. First, the random pattern
hypothesis (null hypothesis), which suggests that road-kill and passage-use patterns are a result of
probabilities that are not influenced by spatial attributes or context. In contrast to the random pattern
hypothesis, I also examined several ecologically meaningful hypotheses: 1) The local scale hypothesis
suggested that wildlife road-kill and passage-use patterns were spatially aggregated due to local roadsection attributes; 2) The landscape scale hypothesis stated that road-kill and passage-use patterns were
influenced by landscape attributes; 3) The scale dependent hypothesis proposed that both local and
landscape variables interact to determine road-kill and passage-use patterns.
First, I tested the random pattern hypothesis (null hypothesis). The analyses of all of the
examined effects of spatial determinants on passage-use rates by wildlife showed the null model was the
best predictive one in only two out of 16 GLM sets (for local scale attributes for prey species, and for
landscape scale attributes at observation radius of 2000 m for predator species). Additionally, passages
64
significantly differed from one another in terms of predation occurrence and frequency (p<<0.01).
Overall, these results indicate that both passage-use patterns I examined are not random.
As for the pattern of road-kills, I examined the spatial distribution of road-kill locations. I found the
Average Nearest Neighbor Index (ANNI) for all the examined roads to be significantly different from the
expected value for random or over-dispersed patterns. This result was consistent for all the examined
taxonomic groups (birds, reptiles and mammals). Consequently, these results provide evidence that roadkills are not randomly distributed throughout all the examined roads, and for all of the examined
taxonomic groups. Moreover, based on the Ripley's K analysis I can conclude that road-kills of all the
examined taxonomic groups are spatially aggregated in all three roads.
However, as can be seen in figures 3.1.1.2, 3.1.2.1 and 3.1.3.1, the spatial range at which these
aggregations are found is up to an observation range of ~400 m in roads ‘44’ and ‘424’, in comparison
with an observation range of up to ~1000 m in road ‘3’. This implies that road ‘3’ imposes a more
profound fragmentation on the landscape, funneling movement of wildlife to more specific and narrower
locations. This observation consists with road ‘3’ being the most intense road in this study, both in means
of average traffic volume, average traffic speed, number of lanes and the presence of 'Jersey barriers'. It
seems that wildlife animals are more deterred from road '3' due to its higher intensity and less keen to
cross it. In turn, they are funneled to more definite locations than in roads '44' and '424'. The latter
explanation is supported by Coelho et al. (2008), which found an amplitude of scales showing road-kill
clustering. This amplitude was similar among taxonomic groups on the same road and broader on road
'RS-389' (except for birds) than on 'BR-101'. Similarly, Clevenger et al. (2003) also observed resembling
amplitudes of scales for clusters of mammals and birds within a road, and variations between different
roads in the Bow River Valley, Canada. In line with my conclusion, they suggested that the dissimilarity
in spatial aggregation patterns between roads might result from differences in design and traffic flow.
In accordance with the local scale hypothesis, I examined the effect of attributes that define the local
scale spatial context of road-sections and passages. As for road-kill patterns, I found that the road-surface
conjunction correlated with rates of road-kills. Road-sections that were aligned with their surroundings
exhibited higher rates of road-kills for all of the examined taxonomic groups. Likewise, the distance to
the nearest road-crossing passage was also important in determining road-kills of all of the examined
taxonomic groups. For mammals, I also found that there were higher numbers of road-kills on
illuminated road-sections.
Taken together, these results indicate that the local-scale spatial context of road-sections determines,
at least partly, the existing patterns of road-kills. It seems that wild animals are more likely to cross the
road at localities where the road is even with its surroundings. These results are consistent with findings
by Clevenger et al. (2003), showing that collisions with animals are rare where roadsides have high
65
embankments. They conclude that optimum crossing points for animals, and in turn collisions, often
concentrate in areas where roads run level with the adjacent landscape. Similar results by Malo et al.
(2004) propose that embankments should be at least 2 m high to discourage animals from crossing.
At the local scale, I also found a negative correlation between road-kill rates and the distance to the
nearest passage. This means that proximity to crossing passages increases the likeliness of wild animals
to cross the road. These results are supported by a recent study by Grilo et al. (2009), concluding that
although passages are commonly suggested as a mitigation measure to prevent road casualties (Forman et
al. 2003, Iuell et al. 2003, Bissonette and Adair 2008), surprisingly, passages in their study were related
with red fox road mortality. They suggest that that number of passages (or distance to them) per-se is
probably not directly related to road-kills but rather that their proximity to good habitat exposes animals
using these habitats to road danger. In contrast to these conclusions, some studies showed that the
presence of passages (i.e. underpasses) reduces collision rates (Foster and Humphrey 1995, Clevenger and
Waltho 2000, Seiler 2003). I suggest that a solution to this contradiction may be in the identity of the
examined species. Clearly, this observation requires further investigation and is worthy to focus on.
My results also suggest that mammals are more prone to road-kills in illuminated road-sections. It is
possible that illumination masks vehicle lights and as such makes it harder for mammals to spot them.
Another explanation may be that artificial illumination is usually correlated with proximity to human
settlements and is located mainly at junctions, two factors which are known to be correlated with high
rates of road-kills (Malo et al. 2004, Grilo et al. 2009). It is also possible that light attracts some wildlife
species to cross roads. If true, then artificial illumination at specific road-sections might act as an
ecological trap.
As for passage-use patterns, I explored the effect of attributes that describe the local-scale spatial
context of crossing passages, on passage usage rates by wildlife species and on predation pressure in
passage vicinity. My results revealed that local-scale attributes -- including the passage's physical
characteristics (i.e. openness and height from road), the level of human use of the passage and plant cover
– had significant effects. Passage-use rates of predators, as well as predation occurrence, were both
negatively correlated with human-use levels. This means that predators avoid passages that are often used
by humans. Moreover, the positive correlation I found between passage openness and passage usage
indicates that predators prefer to cross at passages that are more open, in which visibility of the other side
is high.
Another pattern which I found was the positive correlation between the passage-use rates of prey
species, as well as predation occurrence, and the relative plant cover at the passage. This means that prey
species prefer to cross through passages with dense plant cover. Clearly, dense areas increase antipredation mechanisms for prey species. The plant cover in these dense areas offers safety by providing
66
cover and hiding places (Lima 1995, Abramsky et al. 2002, Creel 2005, Wirsing et al. 2007). On the
other hand, this preference by prey species might attract predators to use these environments to catch
prey. Furthermore, high plant cover may provide better prey catching opportunities through hiding places
and ambushing success. This observation is supported by studies that provide evidence that crossing
passages are used by predators to capture prey (Hunt et al. 1987, Foster and Humphrey 1995; see review
by Little et al. 2002). I further suggest that passages may serve as bottlenecks for wildlife movement and
could potentially increase prey vulnerability to predation. Such passages are generally exposed,
restricted, and often narrow sites (Reed et al. 1975, Yanes et al. 1995, Clevenger et al. 2001a). As such,
these passages reduce the effectiveness of mechanisms used by prey species to avoid detection or escape
predators.
With respect to all of the above, I propose that predators prefer more open and exposed environments
for crossing passages whereas prey species prefer more enclosed and concealed crossing passages. This
conclusion is in line with fundamental mechanisms of predator-prey interactions. As well documented
(Sih 1980, Edwards 1983, Stephens and Peterson 1984, Sweitzer 1996, Gilliam and Fraser 1987,
Altendorf et al. 2001, Hernández and Laundré 2005, Ripple and Beschta 2004, Bergman et al. 2006), prey
species wish to maximize anti-predation mechanisms, even at the price of losing feeding opportunities.
Thus, I argue that predators choose their preferred crossing passages and use them more frequently. In
turn, prey species face predation risk at passages, which leads them to prefer passages which offer more
cover and protection.
These conclusions are consistent with previous observations. Studies showed that animals avoided
the proximity of humans at points where they crossed roads, preferring to approach roads hidden by tree
and shrub cover (Bashore et al. 1985, Jaren et al. 1991, Clevenger et al. 2003, Seiler 2003). Further
support is found in studies showing that the dimensions of passages were also important for crossing rates
of vertebrates (Ulbrich 1984, Ballon 1985 [as cited in Yanes et al. 1995]). The size and shape of a
particular crossing passage may affect crossing success (Reed et al. 1975, Cain et al. 2003, Clevenger and
Waltho 2005). In addition, previous studies also support my conclusions regarding passage openness and
the difference between predator and prey species. These studies have shown that for some species, the
relative openness in a passage was more important than the overall size (Foster and Humphrey 1995,
Clevenger and Waltho 2005). Crossing passages along the Trans-Canada Highway with high openness
ratios (i.e. short in length, high and wide) were used more intensively by grizzly bears, wolves, elk, and
deer, whereas more constrictive passages were used more frequently by black bears and cougars
(Clevenger and Waltho 2005). Moreover, Rosell et al. (1997) showed that tunnels that allowed animals to
see the other end were positively correlated with their use by some species. Conversely, some studies
(Rodriguez et al. 1996, Clevenger and Waltho 1999) have suggested that smaller passages might be better
67
for some small mammals. Lastly, there is some supporting evidence that predators use crossing structures
to increase prey capture (Hunt et al. 1987, Foster and Humphrey 1995), which consequently can reduce
the use of crossing structures by prey species. Similar to my findings, studies have suggested that
passages that were exposed, restricted, or narrow reduced the effectiveness of escape mechanisms of prey
species (Reed et al. 1975, Yanes et al. 1995, Clevenger et al. 2001a).
To summarize, I have shown that factors describing the spatial context of road-sections at a local
scale, such as illumination, distance to nearest crossing passage and the road-landscape conjunction,
clearly affect road-kill rates. Similarly, factors that relate to the local-scale spatial context of crossing
passages evidently affect passage-use patterns. As a result, I conclude that the local scale hypothesis is
highly supported by my results of the local scale attribute effects.
Further analyses also support the landscape hypothesis. I found that different land-use covers
influenced both permeability patterns (i.e. road-kills and passage-use), for all of the examined taxonomic
groups. First, bird road-kills were positively correlated with dwarf-shrub steppe cover. Likewise, I found
a positive correlation between road-kills of reptiles and relative cover of perennial agriculture. I suggest
that these land-uses (dwarf-shrub steppe and perennial agriculture) serve as potential sources of resources
for the examined species (birds and reptiles, respectively). As such, these land-uses might act as drawing
forces for bird and reptile movement, increasing the probability of crossing a nearby road, and, in turn,
also increasing rates of road-kills.
My analysis also revealed that rates of passage-use by predators were positively correlated with the
relative cover of dwarf-shrub steppe and natural forest within an observation radius of 250 m. This means
that more cover of these land-uses within 250 m radius results with higher rates of passage-use by
predators. As before, I suggest that these land-uses offer predator species resources and suited habitats
near passages, and as such, increase the usage of these passages by these species.
Next, I showed that the relative cover of natural forests is negatively correlated with road-kill
numbers of all groups. I suggest that natural forests serve as good habitats, which may provide most of
the species’ needs, due to the resources and cover they provide (Malo et al. 2004). In turn, high cover of
natural forests might mean that the animals do not require long-distance movement across the landscape
in order to meet basic needs, thus keeping them away from crossing roads. Interestingly, there was a
positive correlation between the cover of natural forests and patterns of passage-use. It seems that when
located near passages, natural forests increase rates of passage-use by wildlife. I suggest that both of
these patterns can be explained by the habitat suitability of natural forests. The better the habitat, the
more individuals and species, and therefore more passes in nearby passages. However, as natural forests
are a stable environment, the species inhabiting it avoid disruptions, such as roads (Fahrig 2007).
Therefore, these animals avoid crossing the road-surface. In turn, this might explain the lower numbers
68
of road-kills observed at road-sections near natural forests. I also found a negative correlation between
predator passage-use rates and human planted forest or annual agriculture covers (at 500 m and 100 m
radius, respectively). As for prey species, I found that rates of passage use were negatively affected by
annual agriculture and human settlements (both at 250 m radius).
To summarize, I showed that factors describing the spatial context of road-sections and passages at a
landscape scale (i.e. the matrix of land-uses), clearly affect the related road permeability patterns (i.e.
road-kills and passage-use). I argue that this matrix of land-uses reflects the distribution of resources, and
as such determines species abundance. Moreover, the matrix of land-uses determines main movement
paths as different land-uses serve as the attracting or deterring forces for wildlife movement.
My conclusions are strongly supported by various studies. For example, Malo et al. (2004) conclude
that road-kill clustering at a landscape scale is mainly associated with forest areas, relatively low crop
cover, a fairly high habitat diversity and low presence of buildings. They argue that the higher collision
hazard is not in forested areas, but when dense forest cover is found side-by-side with open habitats.
Moreover, the existence of areas with frequent road-kills at zones with dense tree cover and a minimal
presence of buildings is relatively common. Another example relates to finding that the majority of large
herbivores involved in collisions in Europe and North America live in forest habitats and avoid areas with
high human activity (Bashore et al. 1985, Finder et al. 1999, Hubbard et al. 2000, Nielsen et al. 2003). In
agreement, Allen and McCullough (1976) concluded that large variability in road-kill rates exists among
different road sections because animals use a variety of habitat types. In Spain, red deer, roe deer and
wild boar are forest species, but they forage largely in open areas, especially in the absence of human
disturbance (Ballesteros 1998, Blanco 1998; [as cited in Malo et al. 2004]). The same was concluded for
species involved in collisions in North America, such as the white-tailed deer (Finder et al. 1999,
Hubbard et al. 2000, Nielsen et al. 2003). The habitat diversity index was the most useful variable for
detecting road sections with high collision rates in Illinois, Iowa and Minnesota in the USA. Collisions
tended to occur in areas with some herbaceous cover (grass and crops) mixed with forest (Finder et al.
1999, Hubbard et al. 2000, Nielsen et al. 2003). Similarly, in Pennsylvania, wildlife road-kill rates were
high in areas where deer leave the forest to feed in open patches (Puglisi et al. 1974).
I also compared importance index values and correlation direction for each examined land-use across
observation radii. Here I found that values of importance index change over observation radii. These
results are consistent with spatial ecology theory (Turnner 2005b), as fine-tuning for the proper
observation radius should uncover the range at which each land-use has the most profound effect. Less
intuitively, the correlation pattern was found to be contrary in different radii of observation. I suggest that
utilizing of different land-uses by wildlife species in different area sizes (i.e. observation radii) may
explain this apparent contradiction. It is possible that some land-uses may affect wildlife movement
69
differently at different radii of observation. For example, I found that relative cover of human settlements
around passages was negatively correlated with passage-use rates of predators at small radii but positively
correlated at large radii. I suggest that high human presence at small distances from passages discourages
wildlife movement due to human disturbance (Bashore et al. 1985, Jaren et al. 1991, Clevenger et al.
2003, Seiler 2003, Malo et al. 2004), but attracts wildlife if present at large distances due to resource
availability (Fedriani et al. 2001, Admasu et al. 2004, Contesse et al. 2004, Baghli and Verhagen 2005,
Kusak et al. 2005).
The results regarding the landscape-scale effects support the landscape scale hypothesis. In
accordance with the specific findings, I argue that the landscape matrix influences movement pattern and
funnels movement of wildlife to specific road sections and/or crossing passages. In agreement with
previous studies, I claim that patch composition and physiognomy (the position of the land-use and
habitat patches in the landscape) determine species movement patterns and activity levels in different
patches (Dunning et al., 1992, Wiens et al., 1993, Forman 1995, Turner et al., 2001). More specifically, I
suggest that different land-uses act as the attracting forces for wildlife movement. Thus, the distribution
of resources determine movement flux in specific locations (road-sections and passages), according to
their spatial context. In turn, this movement flux influences patterns of road-kills and passage-use. Put
simply, I claim that higher rates of wildlife movement near roads results in more attempts of road
crossings and more events of collisions with vehicles (i.e. road-kills). Likewise, higher activity levels
near passages increases rates of use by wildlife. This conclusion is consistent with a review by Coffin
(2007) concluding that the determinants for animal road-kills are driven mostly by the spatial arrangement
of resources. Coffin further supports my conclusion by suggesting that animals get hit by vehicles and
die while trying to reach resources (food, water, den sites, etc.).
All these results and conclusions show that both local-scale spatial context as well as the landscape
scale context influence road permeability patterns. This means that, in line with the scale dependent
hypothesis, processes affecting overall wildlife road-kills and passage uses depend on the complex
relationship between factors at different spatial scales. Furthermore, my analyses of models including
variables from different scales emphasize the importance of taking into account both scales. This
conclusion is supported by recent studies. Beaudry et al. (2008) conclude that conservation interventions
are most likely to be effective in mitigating the effects of road mortality when implemented at the road
segment and population scales. Further support exists in a study by Serrano et al. (2002). They conclude
from their two-scale study (regional and local) of road-induced fragmentation that examining both scales
proves very useful for improving landscape management and mitigation regimes. A study by Grilo et al.
(2009) also supports the importance of including spatial factors from both the local and the landscape
scale in models that evaluate road-kill patterns. They conclude that variables from different scales (e.g.
70
favorable habitat, curves in the road, and low human disturbance) serve as major contributors to the
deadliest road segments.
Interestingly, my robust analyses highlight the relatively higher importance of landscape scale
variables. In almost all cases landscape scale variables had the highest values of importance index. I
claim that this emphasizes the status of landscape scale variables in determining models' performance in
predicting road permeability patterns. This conclusion agrees well with the scale hierarchy theory
(Turner 2005b). First, an organism needs to encounter a road, and only then does it decide whether to
cross it or not, and in which manner. Accordingly, I claim that permeability patterns are first influenced
by the landscape scale spatial context of each road-section. This landscape scale context of the roadsection determines the species' density and main movement paths and in turn, the movement flux. Then,
local scale spatial determinants affect the decision of an individual regarding road-crossing options, thus
influencing the result of crossing events within the specific road-section.
For example, I found that landscape scale variables had higher relative importance on predation
occurrence. As such, my results imply that the landscape scale has a more profound effect on predation
pressure at the vicinity of crossing passages. Clearly, one can deduce that in the absence of a predator,
the local scale attributes, which affect the daring of the predator (through interference), do not come into
play. Accordingly, I suggest that landscape-scale variables define movement pathways of predators and
as such define movement flux in certain locations. Within these locations, local-scale variables such as
plant cover, affect animals’ decisions whether to use a particular crossing passage.
The negative correlation between the change in predation frequency and the distance between
passages (figure 3.3.1) is somewhat counter-intuitive. I suggest that this can be explained by the
overlapping of predator activity between close-by passages. Assuming that predators utilize not just the
structure itself but also its surrounding, when passages are very close one to each other, there is a possible
overlap in the activity of predators in these areas. This overlap changes the relative effect of predation
and may result in synergetic pressures on predator-prey interactions. Hence, strategies of placing
passages for conserving focal animals need to consider the pressure imposed by their predators.
If we take into consideration that building passages to provide full permeability at each point is
practically unrealistic, this finding has two main consequences. The first, suggests that passages can be
built nearby each other if the same predator individuals use both of them. In this case, the predator uses
the passages alternatively. As such it reduces the use rate of each one of them and thus reduces the
predation pressure at the passage itself. The second, less intuitive, suggests that it might be better for
passages not to be too close one to another in order not to create high predation pressure zones (between
the passages).
71
Moreover, while trying to attract and guide wildlife to a crossing passage, prey individuals might be
funneled to high predation pressure locations. In turn, high level of predation pressure may force prey to
avoid the passage and cross on the road surface. At locations with high risk for road-kills, this may result
in turning these locations into ecological traps for prey species. An indication for this potential effect
comes from the higher numbers of road-kills in the vicinity of crossing passages found in this work (i.e.
the negative correlation between road-kill rates and distance to the nearest passage; p<0.001 for all of the
examined taxonomic groups).
In general, the results of this thesis suggest that road permeability is strongly affected by many scaledependent ecological processes, as well as by multi-scale processes. The results also provide evidence
that the influence of predator-prey interactions in the vicinity of roads might result in passages becoming
ecological traps. Although there are still many gaps of knowledge in our understanding of wildlife
conservation in road-oriented landscapes, my study demonstrates the benefit of a scientifically-based
examination of road permeability. Based on the above findings I suggest that mitigation practices should
take into account various spatial scale-dependent determinants and in turn, should consider their related
ecological processes. Doing so can lead to efficient and realistic practices to protect species diversity and
stabilize ecological systems in road-dominated areas. In line with conclusions from contemporary
studies, it is highly emphasized that mitigation practices should set quantitative rather than qualitative
goals (e.g. 4 road-kill events per year, 7 out of 9 individuals crossing at a certain passage; as opposed to
reducing road-kill rates and/or improving road permeability). It is subsequently possible to determine
whether such goals have been met (van der Ree et al. 2007, van der Grift and Schippers 2013).
With respect to the multi-scale effects, my thesis suggests, for the first time as far as I know, an
explicit ecological mechanism for road permeability patterns -- I relate landscape scale to the overall
movement flux in a certain location. Whereas, at a local scale, when an individual meets the road-section,
I claim local heterogeneity affects its road-crossing decision. I suggest that the predictive power of
models related to road permeability is enhanced by linking road permeability patterns to explanatory
variables through ecological meaningful processes. Moreover, one can relate between the mitigation
goals and practical actions, and as a result increase the probability of success. Lastly, I argue that these
conclusions highlight the complexity of road-induced fragmentation and its solutions for wildlife
conservation. Future studies should examine the suggested mechanism for road permeability patterns.
There is still a need to reveal the explicit links between the processes operating at different scales and the
way various groups of organisms respond to these scale-dependent processes under different landscape
heterogeneities. Moreover, I suggest that future studies should focus on examining the effect of
heterogeneity per-se (e.g., using advanced landscape configuration indices, such as fractal geometry and
percolation matrices) on road permeability patterns. I believe that such future directions will contribute to
72
our conservation objectives by better understanding of passage functionality, contribution to animal
survival, and avoidance of ecological traps.
73
References
Abramsky, Z., M. L. Rosenzweig, and A. Subach. 2002. The costs of apprehensive foraging.
Ecology 83:1330–1340.
Addicott, J. F., J. M. Aho, M. F. Antolin, D. K. Padilla, J. S. Richardson, and D. A. Soluk. 1987.
Ecological neighborhoods: scaling environmental patterns. kos 49:340-346.
Admasu, E., S. J. Thirgood, A. Bekele, and M. Karen Laurenson. 2004. Spatial ecology of golden
jackal in farmland in the Ethiopian Highlands. African Journal of Ecology 42:144-152.
Alexander, S. M., N. M. Waters, and P. C. Paquet. 2005. Traffic volume and highway permeability
for a mammalian community in the Canadian Rocky Mountains. Canadian GeographerGeographe Canadien 49:321-331.
Allen, R. E., and D. R. McCullough. 1976. Deer-car accidents in southern Michigan. Journal of
Wildlife Management 40:317-325.
Altendorf, K. B., J. W. Laundré, C. A. López González, and J. S. Brown. 2001. Assessing effects of
predation risk on foraging behavior of mule deer. Journal of Mammalogy 82:430-439.
Anderson, D. R., and K. P. Burnham. 2002. Avoiding pitfalls when using information-theoretic
methods. Journal of Wildlife Management 66:912-918.
Anderson, D. R., K. P. Burnham, and W. L. Thompson. 2000. Null hypothesis testing: Problems,
prevalence, and an alternative. Journal of Wildlife Management 64:912-923.
Aresco, M. J. 2005. Mitigation measures to reduce highway mortality of turtles and other
herpetofauna at a north Florida lake. Journal of Wildlife Management 69:549-560.
Baghli, A., and R. Verhagen. 2005. Activity patterns and use of resting sites by polecats in an
endangered population. Mammalia mamm 69:211-222.
Bailey, T. C., and A. C. Gatrell 1995. Interactive Spatial Data Analysis. Longman Scientific &
Technical, Harlow, U.K.
Barnett, J. L., R. A. How, and W. F. Humphreys. 1978. Use of Habitat Components by Small
Mammals in Eastern Australia. Australian Journal of Ecology 3:277-285.
Bashore, T. L., W. M. Tzilkowski, and E. D. Bellis. 1985. Analysis of Deer-Vehicle Collision Sites in
Pennsylvania. Journal of Wildlife Management 49:769-774.
Beaudry, F., P. G. Demaynadier, and M. L. Hunter. 2008. Identifying road mortality threat at
multiple spatial scales for semi-aquatic turtles. Biological Conservation 141:2550-2563.
74
Bellis, M. A., S. D. Jackson, C. R. Griffin, P. S. Warren, and A. O. and Thompson. 2007. Utilizing a
Multi-Technique, Multi-Taxa Approach to Monitoring Wildlife Passageways on the Bennington
Bypass in Southern Vermont. eScholarship, UC Davis: Road Ecology Center.
Bergman, E. J., R. A. Garrott, S. Creel, J. J. Borkowski, R. Jaffe, and F. Watson. 2006. Assessment
of prey vulnerability through analysis of wolf movements and kill sites. Ecological Applications
16:273-284.
Bissonette, J. A., and W. Adair. 2008. Restoring habitat permeability to roaded landscapes with
isometrically-scaled wildlife crossings. Biological Conservation 141:482-488.
Bolger, D. T., T. A. Scott, and J. T. Rotenberry. 2001. Use of corridor-like landscape structures by
bird and small mammal species. Biological Conservation 102:213-224.
Bruinderink, G. W. T. A. G., and E. Hazebroek. 1996. Ungulate Traffic Collisions in Europe
Colisiones de Trápico con Ungulados en Europa. Conservation Biology 10:1059-1067.
Buckland, S. T., K. P. Burnham, and N. H. Augustin. 1997. Model selection: An integral part of
inference. Biometrics 53:603-618.
Cain, A. T., V. R. Tuovila, D. G. Hewitt, and M. E. Tewes. 2003. Effects of a highway and
mitigation projects on bobcats in Southern Texas. Biological Conservation 114:189-197.
Carr, L. W., and L. Fahrig. 2001. Effect of road traffic on two amphibian species of differing
vagility. Conservation Biology 15:1071-1078.
Clarke, G. P., P. C. L. White, and S. Harris. 1998. Effects of roads on badger Meles meles
populations in south-west England. Biological Conservation 86:117-124.
Clevenger, A. P., B. Chruszcz, and K. Gunson. 2001a. Drainage culverts as habitat linkages and
factors affecting passage by mammals. Journal of Applied Ecology 38:1340-1349.
Clevenger, A. P., B. Chruszcz, and K. E. Gunson. 2001b. Highway mitigation fencing reduces
wildlife-vehicle collisions. Wildlife Society Bulletin 29:646-653.
Clevenger, A. P., B. Chruszczc, and K. E. Gunson. 2003. Spatial patterns and factors influencing
small vertebrate fauna road-kill aggregations. Biological Conservation 109:15-26.
Clevenger, A. P., and N. Waltho. 2000. Factors influencing the effectiveness of wildlife
underpasses in Banff National Park, Alberta, Canada. Conservation Biology 14:47-56.
Clevenger, A. P., and N. Waltho. 2005. Performance indices to identify attributes of highway
crossing structures facilitating movement of large mammals. Biological Conservation 121:453464.
75
Coelho, I. P., A. Kindel, and A. V. P. Coelho. 2008. Roadkills of vertebrate species on two
highways through the Atlantic Forest Biosphere Reserve, southern Brazil. European Journal of
Wildlife Research 54:689-699.
Coffin, A. W. 2007. From roadkill to road ecology: A review of the ecological effects of roads.
Journal of Transport Geography 15:396-406.
Contesse, P., D. Hegglin, S. Gloor, F. Bontadina, and P. Deplazes. The diet of urban foxes (Vulpes
vulpes) and the availability of anthropogenic food in the city of Zurich, Switzerland.
Cooper, C. B., J. R. Walters, and H. Ford. 2002. Effects of remnant size and connectivity on the
response of Brown Treecreepers to habitat fragmentation. Emu 102:249-256.
Corlatti, L., K. Hacklaender, and F. Frey-Roos. 2009. Ability of Wildlife Overpasses to Provide
Connectivity and Prevent Genetic Isolation. Conservation Biology 23:548-556.
Creel, S., J. A. Winnie, B. Maxwell, K. Hamlin, and M. Creel. 2005. Elk alter habitat selection as an
antipredator response to wolves. Ecology 86:3387–3397.
Cressie, N. 1991. Statistics for spatial data. John Wiley and Sons New York.
Dunning, J. B., B. J. Danielson, and H. R. Pulliam. 1992. Ecological Processes That Affect
Populations in Complex Landscapes. Oikos 65:169-175.
Ebdon, D. 1985. Statistics in Geography Blackwell.
Edwards, J. 1983. Diet shifts in moose due to predator avoidance. Oecologia 60:185-189.
Fahrig, L. 2007. Non‐optimal animal movement in human‐altered landscapes. Functional Ecology
21:1003-1015.
Fahrig, L., and G. Merriam. 1994. Conservation of Fragmented Populations. Conservation Biology
8:50-59.
Fedriani, J. M., T. K. Fuller, and R. M. Sauvajot. 2001. Does availability of anthropogenic food
enhance densities of omnivorous mammals? An example with coyotes in southern California.
Ecography 24:325-331.
Fenn, M. G. P., and D. W. Macdonald. 1995. Use of middens by red foxes: risk reverses rhythms
of rats. Journal of Mammalogy 76:130–136.
Ferreras, P., J. J. Aldama, J. F. Beltran, and M. Delibes. 1992. Rates and Causes of Mortality in a
Fragmented Population of Iberian Lynx Felis-Pardina Temminck, 1824. Biological Conservation
61:197-202.
Finder, R. A., J. L. Roseberry, and A. Woolf. 1999. Site and landscape conditions at white-tailed
deer/vehicle collision locations in Illinois. Landscape and Urban Planning 44:77-85.
76
Forman, H., and F. Merrick. 2003. Discipline - Learning the rules of the management high road.
Journal of Nursing Administration 33:65-67.
Forman, R. T. 1995. Land mosaics: the ecology of landscapes and regions. Cambridge university
press.
Forman, R. T. T., and L. E. Alexander. 1998. Roads and their major ecological effects. Annual
Review of Ecology and Systematics 29:207.+Forman, R. T. T., B. Reineking, and A. M. Hersperger. 2002. Road traffic and nearby grassland
bird patterns in a suburbanizing landscape. Environmental Management 29:782-800.
Forman, R. T. T., Sperling, D., Bissonette, J.A., Clevenger, A.P., Cutshall, C.D., Dale, V.H., Fahrig,
L., France, R., Goldman, C.R., Heanue, K., Jones, J.A., Swanson, F.J., Turrentine, T., Winter, T.C.
2003. Road ecology: science and solutions. Island Press, Washington.
Fortin, M. J., and M. T. T. Dale 2005. Spatial Analysis: A Guide for Ecologists. Cambridge
University Press, Cambridge.
Foster, M. L., and S. R. Humphrey. 1995. Use of Highway Underpasses by Florida Panthers and
Other Wildlife. Wildlife Society Bulletin 23:95-100.
Getis, A. 1984. Interaction Modeling Using 2nd-Order Analysis. Environment and Planning A
16:173-183.
Gibeau, M. L., A. P. Clevenger, S. Herrero, and J. Wierzchowski. 2002. Grizzly bear response to
human development and activities in the Bow River Watershed, Alberta, Canada. Biological
Conservation 103:227-236.
Gilliam, J. F., and D. F. Fraser. 1987. Habitat selection under predation hazard: test of a model
with foraging minnows. Ecology 68:1856-1862.
Grilo, C., J. A. Bissonette, and M. Santos-Reis. 2009. Spatial-temporal patterns in Mediterranean
carnivore road casualties: Consequences for mitigation. Biological Conservation 142:301-313.
Haining, R. 2003. Spatial Data Analysis: Theory and Practice. Cambrideg University Press,
Cambridge.
Haskell, D. G. 2000. Effects of forest roads on macroinvertebrate soil fauna of the Southern
Appalachian mountains. Conservation Biology 14:57-63.
Haxton, T. 2000. Road mortality of snapping turtles, Chelydra serpentina, in central Ontario
during their nesting period. Canadian Field-Naturalist 114:106-110.
Hayne, D. W. 1949. Calculation of Size of Home Range. Journal of Mammalogy 30:1-18.
77
Heilman, G. E., J. R. Strittholt, N. C. Slosser, and D. A. Dellasala. 2002. Forest fragmentation of
the conterminous United States: Assessing forest intactness through road density and spatial
characteristics. Bioscience 52:411-422.
Hernández, L., and J. W. Laundré. 2005. Foraging in the'landscape of fear'and its implications for
habitat use and diet quality of elk Cervus elaphus and bison Bison bison. Wildlife Biology 11:215220.
Hodson N.L., D. W. S. 1965. The road deaths enquiry. Bird Study 12:90-99.
Hubbard, M. W., B. J. Danielson, and R. A. Schmitz. 2000. Factors influencing the location of
deer-vehicle accidents in Iowa. Journal of Wildlife Management 64:707-713.
Huber, D., J. Kusak, and A. Frkovic 1998. Traffic kills of brown bears in Gorski kotar, Croatia.
International Association Bearresearch & Management, West Glacier.
Hunt, A., Dickens, J., Whelan, R.J. 1987. Movement of mammals through tunnels under railway
lines. Australian Zoologist 24:89–93.
Iuell, B., Bekker, H., Cuperus, R., Dufek, J., Fry,, H. G. L., C., Hlavac, V., Keller, J., Le Marie, B.
Wandall, Rosell Pagès, C., S, and T. angwine, Torslov, N. . 2003. Wildlife and Traffic - A European
Handbook for Identifying Conflicts and Designing Solutions in R. a. H. E. d. Public Works and
Water Management, editor, Delft, The Netherlands.
Jaarsma, C. F., F. van Langevelde, and H. Botma. 2006. Flattened fauna and mitigation: Traffic
victims related to road, traffic, vehicle, and species characteristics. Transportation Research Part
D-Transport and Environment 11:264-276.
Jackson, S. D. 1999. Overview of transportation related wildlife problems. Pages 73-99 in G. P.
Evink G L, and Zeigler D, editor. Third International Conference on Wilflife Ecology and
Transportation. FL-ER.
Jaren, V., R. Andersen, M. Ulleberg, P. H. Pedersen, and B. Wiseth. 1991. Moose-Train Collisions
- the Effects of Vegetation Removal with a Cost-Benefit-Analysis. Alces, Vol 27:93-99.
Joyce, T. L., and S. P. Mahoney. 2001. Spatial and Temporal Distributions of Moose-Vehicle
Collisions in Newfoundland. Wildlife society bulletin 29:281-291.
Kuitunen, M., E. Rossi, and A. Stenroos. 1998. Do highways influence density of land birds?
Environmental Management 22:297-302.
Kusak, J., A. M. Skrbinšek, and D. Huber. 2005. Home ranges, movements, and activity of wolves
(Canis lupus) in the Dalmatian part of Dinarids, Croatia. European Journal of Wildlife Research
51:254-262.
78
Land, D. a. M. L. 1996. Wildlife crossing designs and use by Florida panthers and other wildlife in
southwest Florida. Pages 1-6. International Conference on Wildlife Ecology and Transportation.
Leibold, M. A., M. Holyoak, N. Mouquet, P. Amarasekare, J. Chase, M. Hoopes, R. Holt, J. Shurin,
R. Law, and D. Tilman. 2004. The metacommunity concept: a framework for multi‐scale
community ecology. Ecology letters 7:601-613.
Levin, S. A. 1970. Community equilibria and stability, and an extension of the competitive
exclusion principle. American Naturalist:413-423.
Levin, S. A. 1992. The Problem of Pattern and Scale in Ecology. Ecology 73:1943-1967.
Lima, S. L. 1998. Nonlethal effects in the ecology of predator-prey interactions - What are the
ecological effects of anti-predator decision-making? Bioscience 48:25-34.
Lima, S. L., and L. M. Dill. 1990. Behavioral Decisions Made under the Risk of Predation - a
Review and Prospectus. Canadian Journal of Zoology-Revue Canadienne De Zoologie 68:619640.
Little, S. J., Harcourt, R.G., Clevenger, A.P. 2002. Do wildlife passages act as prey-traps?
Biological Conservation 107:135–145.
Loveridge, A. J., M. Valeix, Z. Davidson, F. Murindagomo, H. Fritz, and D. W. Macdonald. 2009.
Changes in home range size of African lions in relation to pride size and prey biomass in a semiarid savanna. Ecography 32:953-962.
Mader, H. J. 1984. Animal Habitat Isolation by Roads and Agricultural Fields. Biological
Conservation 29:81-96.
Malo, J. E., F. Suarez, and A. Diez. 2004. Can we mitigate animal-vehicle accidents using
predictive models? Journal of Applied Ecology 41:701-710.
Mata, C., I. Hervas, J. Herranz, J. E. Malo, and F. Suarez. 2009. Seasonal changes in wildlife use of
motorway crossing structures and their implication for monitoring programmes. Transportation
Research Part D-Transport and Environment 14:447-452.
McArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Monographs in
population Biology 1:22.
Mitchell, A. 2005. The ESRI Guide to GIS Analysis. ESRI Press.
Ng, S. J., J. W. Dole, R. M. Sauvajot, S. P. D. Riley, and T. J. Valone. 2004. Use of highway
undercrossings by wildlife in southern California. Biological Conservation 115:499-507.
Nielsen, C. K., R. G. Anderson, and M. D. Grund. 2003. Landscape influences on deer-vehicle
accident areas in an urban environment. The Journal of wildlife management 67:46-51.
79
Norman, T., Lean, B. and A. Finnegan. 1998. The role of fauna underpasses in New South Wales.
Pages 195-207. International Conference on Wildlife Ecology and Transportation.
Noss, R. F., H. B. Quigley, M. G. Hornocker, T. Merrill, and P. C. Paquet. 1996. Conservation
biology and carnivore conservation in the Rocky Mountains. Conservation Biology 10:949-963.
Oehler, J. D., and J. A. Litvaitis. 1996. The role of spatial scale in understanding responses of
medium-sized carnivores to forest fragmentation. Canadian Journal of Zoology-Revue
Canadienne De Zoologie 74:2070-2079.
Oxley, D. J., M. B. Fenton, and G. R. Carmody. 1974. Effects of Roads on Populations of Small
Mammals. Journal of Applied Ecology 11:51-59.
Pfister, H. P., Keller, V., Reck, H., Georgii, B. 1997. Bio-Ecological Effectiveness of Wildlife
Overpasses or “Green Bridges” Over Roads and Railway Lines. Herausgegeben vom
Bundesministerium fur Verkehr Abeteilung Strassenbau. Bonn-Bad Godesberg, Germany.
Philcox, C. K., A. L. Grogan, and D. W. Macdonald. 1999. Patterns of otter Lutra lutra road
mortality in Britain. Journal of Applied Ecology 36:748-762.
Pickett, S. T. A., and M. L. Cadenasso. 1995. Landscape Ecology - Spatial Heterogeneity in
Ecological-Systems. Science 269:331-334.
Puglisi, M. J., J. S. Lindzey, and E. D. Bellis. 1974. Factors Associated with Highway Mortality of
White-Tailed Deer. Journal of wildlife management 38:799-807.
Ramp, D., V. K. Wilson, and D. B. Croft. 2006. Assessing the impacts of roads in peri-urban
reserves: Road-based fatalities and road usage by wildlife in the Royal National Park, New South
Wales, Australia. Biological Conservation 129:348-359.
Reed, D. F., Ward, A.L. 1985. Efficacy of methods advocated to reduce deer-vehicle accidents:
research and rationale in the USA. Routes et faune sauvage. Service d’Etudes Techniques de
Routes et Autoroutes, Bagneaux, France:285–293.
Reed, D. F., T. N. Woodard, and T. M. Pojar. 1975. Behavioral Response of Mule Deer to a
Highway Underpass. Journal of Wildlife Management 39:361-367.
Riitters, K., J. Wickham, and J. Coulston. 2004. Use of road maps in national assessments of
forest fragmentation in the United States. Ecology and Society 9.
Ripley, B. 1981. Spatial statistics John Wiley and Sons, New York.
Ripple, W. J., and R. L. Beschta. 2004. Wolves and the ecology of fear: can predation risk
structure ecosystems? BioScience 54:755–766.
Risser, P. G., Karr JR, and R. T. T. Forman 1984. Landscape ecology: directions and approaches.
Nat. Hist. Surv., Champaign.
80
Rodriguez, A., G. Crema, and M. Delibes. 1996. Use of non-wildlife passages across a high speed
railway by terrestrial vertebrates. Journal of Applied Ecology 33:1527-1540.
Rodriguez, A., Crema, G., Delibes, M. 1996. Use of non-wildlife passages across a high-speed
railway by terrestrial vertebrates. Journal of Applied Ecology 33:1527–1540.
Rosell, C., J. Parpal, R. Campeny, S. Jove’, A. Pasquina and J.M. Velasco. 1997 Mitigation of
barrier effect of linear infrastructures on wildlife. Pages 367-372 in K. Canters, editor.
International conference on habitat fragmentation, infrastructure and the role of ecological
engineering. Ministry of Transport, Public Works and Water Management, Delft, The
Netherlands.
Rosen, P. C., and C. H. Lowe. 1994. Highway mortality of snakes in the sonoran desert of
southern Arizona. Biological Conservation 68:143-148.
Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991. Biological Consequences of Ecosystem
Fragmentation - a Review. Conservation Biology 5:18-32.
Seiler, A. 2003. The toll of the automobile.
Seiler, A., and A. Sjolund. 2005. Target orientation for ecologically sound road management in
Sweden. Gaia-Ecological Perspectives for Science and Society 14:178-181.
Serrano, M., L. Sanz, J. Puig, and J. Pons. 2002. Landscape fragmentation caused by the transport
network in Navarra (Spain) - Two-scale analysis and landscape integration assessment.
Landscape and Urban Planning 58:113-123.
Shriver, W. G., T. P. Hodgman, J. P. Gibbs, and P. D. Vickery. 2004. Landscape context influences
salt marsh bird diversity and area requirements in New England. Biological Conservation
119:545-553.
Shuttleworth, C. 2001. Traffic related mortality in a red squirrel (Sciurus vulgaris) population
receiving supplemental feeding. Urban Ecosystems 5:109-118.
Sih, A. 1980. Optimal foraging: partial consumption of prey. American Naturalist 116:281.
Silva, J. d. A., and S. A. Talamoni. 2004. Core area and centre of activity of maned wolves,
Chrysocyon brachyurus (Illiger)(Mammalia, Canidae), submitted to supplemental feeding.
Revista Brasileira de Zoologia 21:391-395.
Slater, F. 2002. An assessment of wildlife road casualties–the potential discrepancy between
numbers counted and numbers killed. Web Ecology 3:33-42.
Soderstrom, B., T. Part, and J. Ryden. 1998. Different nest predator faunas and nest predation
risk on ground and shrub nests at forest ecotones: an experiment and a review. Oecologia
117:108-118.
81
Spellerberg, I. F. 1998. Ecological effects of roads and traffic: a literature review. Global Ecology
and Biogeography 7:317-333.
Stephens, P. W., and R. O. Peterson. 1984. Wolf‐avoidance strategies of moose. Ecography
7:239-244.
Swart, J., and M. J. Lawes. 1996. The effect of habitat patch connectivity on samango monkey
(Cercopithecus mitis) metapopulation persistence. Ecological Modelling 93:57-74.
Sweitzer, R. A. 1996. Predation or starvation: consequences of foraging decisions by porcupines
(Erethizon dorsatum). Journal of Mammalogy:1068-1077.
Taylor, B. D., and R. L. Goldingay. 2004. Wildlife road-kills on three major roads in north-eastern
New South Wales. Wildlife Research 31:83-91.
Team, R. D. C. 2011. R: A Language and Environment for Statistical Computing. R Foundation for
Statistical Computing, Vienna, Austria.
Teixeira, F. Z., I. Pfeifer Coelho, I. Beraldi Esperandio, N. da Rosa Oliveira, F. Porto Peter, S. S
Dornelles, N. Rosa Delazeri, M. Tavares, M. Borges Martins, and A. Kindel. 2013. ARE ROAD-KILL
HOTSPOTS COINCIDENT AMONG DIFFERENT VERTEBRATE GROUPS? Oecologia Australis 17:3647.
Tischendorf, L., and L. Fahrig. 2000. On the usage and measurement of landscape connectivity.
Oikos 90:7-19.
Trombulak, S. C., and C. A. Frissell. 2000. Review of ecological effects of roads on terrestrial and
aquatic communities. Conservation Biology 14:18-30.
Turner, M. G. 1989. Landscape Ecology - the Effect of Pattern on Process. Annual Review of
Ecology and Systematics 20:171-197.
Turner, M. G. 2005a. Landscape ecology in North America: Past, present, and future. Ecology
86:1967-1974.
Turner, M. G. 2005b. Landscape ecology: What is the state of the science? Annual Review of
Ecology Evolution and Systematics 36:319-344.
Turner, M. G., R. H. Gardner, and R. V. O'Neill 2001. Landscape ecology in theory and practicepattern and process. Springer-Verlag Press.
Urban, D. L., R. V. Oneill, and H. H. Shugart. 1987. Landscape Ecology. Bioscience 37:119-127.
van der Grift, E. A., and P. Schippers. WILDLIFE CROSSING STRUCTURES: CAN WE PREDICT
EFFECTS ON POPULATION PERSISTENCE?
82
van der Ree, R., J. A. G. Jaeger, E. A. van der Grift, and A. P. Clevenger. 2011. Effects of Roads
and Traffic on Wildlife Populations and Landscape Function: Road Ecology is Moving toward
Larger Scales. Ecology and Society 16.
van der Ree, R., E. van der Grift, N. Gulle, K. Holland, C. Mata, and F. Suarez. 2007. Overcoming
the barrier effect of roads–how effective are mitigation strategies? An international review of
the use and effectiveness of underpasses and overpasses designed to increase the permeability
of roads for wildlife. Pages 423-431. Proceedings of the 2007 International Conference on
Ecology and Transportation. Center for Transportation and Environment, North Carolina State
University, Raleigh, North Carolina, USA.
van der Zande, A. N., W. J. ter Keurs, and W. J. van der Weijden. 1980. The impact of roads on
the densities of four bird species in an open field habitat—evidence of a long-distance effect.
Biological Conservation 18:299-321.
Wiens, J. A., N. C. Stenseth, B. Van Home, and R. A. Ims. 1993. Ecological mechanisms and
landscape ecology. Oikos 66:369-380.
Wirsing, A. J., M. R. Heithaus, and L. M. Dill. 2007. Living on the edge: dugongs prefer to forage
in microhabitats that allow escape from rather than avoidance of predators. Animal Behaviour
74:93–101.
Yanes, M., J. M. Velasco, and F. Suarez. 1995. Permeability of Roads and Railways to Vertebrates
- the Importance of Culverts. Biological Conservation 71:217-222.
83
‫‪Appendix 1‬‬
‫תאונות דרכים מסכנות חיים בין רכבים לבעלי‪-‬חיים‪ :‬זיהוי מוקדי סיכון ומתן המלצות להפחתת‬
‫שיעור התאונות‬
‫מבוא‪:‬‬
‫מספר תאונות הדרכים בין כלי‪-‬רכב לבעלי חיים נמצא בעליה מתמדת במדינות מפותחות‪ ,‬ביניהן בישראל‪.‬‬
‫מחקרים אחרונים מעריכים כי תאונות אלו הן בסדר גודל של מספר מיליונים ברחבי העולם בכל שנה‪ .‬התוצאה‬
‫של תאונות מסוג זה הינה נזק כלכלי משמעותי וסיכון חיי אדם‪ .‬העלויות החומריות מוערכות בארה"ב במעל‬
‫ל‪ US$0011 -‬לתאונה‪ .‬חשוב מכך‪ ,‬אובדן חיי אדם ופציעות מתרחשות בכ‪ 0% -‬מכלל התאונות המעורבים‬
‫בהן בעלי‪-‬חיים בגודל גוף בינוני‪ ,‬כגון איילים (‪ ,)Conover et al. 1995‬ובכ‪ 01% -‬מהתאונות עם יונקים גדולים‬
‫יותר ( ‪ .)Joyce & Mahoney 2001, Malo et al. 2004‬בנוסף‪ ,‬בעקבות זמינות מוגברת של מזון שמקורו‬
‫מפעילות האדם ('מקור אנטרופוגני')‪ ,‬אוכלוסיות רבות המלוות אדם הגדילו באופן משמעותי את צפיפותן‪ .‬מספר‬
‫פרטים גדול יותר‪ ,‬באזורים מיושבים‪ ,‬מגדיל את הסיכוי למוקדי חיכוך עם בני‪-‬אדם ובתוך כך את הסיכוי‬
‫לתאונות מסכנות חיים‪ .‬חשוב לציין‪ ,‬כי המידע הנ"ל מהווה הערכת חסר מאחר ואינו כולל מידע‪ ,‬אשר בד"כ‬
‫אינו בנמצא‪ ,‬על מספר התאונות הקטלניות שנגרמות עקב בלימת פתע או ניסיון להימנע מפגיעה במקרים של‬
‫חציית בעלי חיים‪.‬‬
‫סיכום תאונות הדרכים בשנים האחרונות בישראל מצביע אף הוא על מגמת גידול בתאונות המערבות בעלי חיים‬
‫(גוטמן ‪ .)0112‬בנוסף‪ ,‬מחקר שהתבצע ע"י ענבר וחבריו (‪ )0110‬ניסה להעריך את מספר התאונות בין יונקים‬
‫יבשתיים לרכבים בארץ‪ .‬בהסתמך על נתונים מחברות הביטוח‪ 200 ,‬תאונות מסוג זה דווחו בין השנים ‪0990-‬‬
‫‪ . 0999‬מאחר וניתן היה לשחזר רק תאונות אשר הובילו לפגיעה פיזית בבני אדם‪ ,‬מציינים החוקרים כי היקף‬
‫התופעה הוא בוודאי גדול בהרבה‪ .‬בנוסף‪ ,‬במהלך ביצוע הסקר‪ ,‬הבחינו החוקרים כי מספר התאונות הכפיל את‬
‫עצמו (‪ . )Inbar et al. 2002‬על פי הידוע לנו‪ ,‬המחקר שמתבצע במעבדתנו הוא היחיד בארץ כיום‪ ,‬שבודק‬
‫באופן מחקרי וסדור תאונות דרכים בין רכבים לבע"ח‪.‬‬
‫העלייה בשיעור התאונות נובעת בעיקר מהתפתחות מערכת הכבישים כתוצאה מגידול באוכלוסיית האדם‪.‬‬
‫למעשה‪ ,‬כבישים מהווים חלק בלתי‪-‬נפרד מהנוף המודרני והפיתוח העולמי‪ .‬אולם‪ ,‬הם גם בעלי השפעות‬
‫הרסניות הן למערכות אקולוגיות‪ ,‬בעיקר בעקבות אובדן וקיטוע בתי גידול‪ ,‬והן לאדם עקב תאונות דרכים‬
‫מסכנות חיים‪ .‬תופעה זו‪ ,‬קיטוע נוף ע"י כבישים‪ ,‬מהווה גורם סיכון רב מעלה לשרידותם של חיות‪-‬בר (שקדי‬
‫‪ ,)Forman & Alexander 1998 ,0112‬ונטל כלכלי כבד לחברה האנושית‪ .‬לחיות הבר‪ ,‬תופעות שליליות אלו‬
‫כוללות‪ ,‬בין השאר‪ ,‬יצירת אוכלוסיות קטנות החשופות לסיכון הכחדה גבוה משני צדי הכביש‪ ,‬ירידה משמעותית‬
‫‪84‬‬
‫בשונות הגנטית של כל אחת מהאוכלוסיות ושינויים בהרכב החברה ויחסי הגומלין עם אוכלוסיות אחרות (כמו‬
‫מתחרים וטורפים)‪ .‬לחברה האנושית הנזק הכלכלי עלול להיות גבוה‪ .‬לשם השוואה‪ ,‬בגרמניה הוא מוערך בכ‪-‬‬
‫‪ $081‬מיליון לשנה (‪ .)Putman 1997‬כבישים גורמים לקיטוע של השטח‪ ,‬מגבילים תנועה של בעלי‪-‬חיים‬
‫ומפחיתים נגישות למשאבים‪ .‬בעקבות קיטוע זה‪ ,‬בעלי חיים נקלעים במהלך תנועתם במרחב לאזורי כבישים‬
‫ומהווים גורם לתאונות דרכים מסכנות חיים‪ .‬לפיכך‪ ,‬יחסי הגומלין של כבישים ומערכות אקולוגיות הינם נושא‬
‫מרכזי במחקר של פיתוח בר‪-‬קיימא (‪.)Alexander et al. 2005‬‬
‫ניתן לסכם‪ ,‬כי קיימת חשיבות עליונה ללימוד שיטות התמודדות עם תאונות הנגרמות מכניסת בעלי חיים‬
‫לתחום הכביש‪ ,‬וכי ישנו אינטרס משותף בהפחתת מספר התאונות המערבות בעלי‪-‬חיים‪ ,‬הן לבטיחות חיי‪-‬‬
‫אדם והן לשמירה על המערכת האקולוגית‪.‬‬
‫מחקר זה שם לעצמו כמטרה לתרום לפתרון צורך זה‪ ,‬ע"י שקלול מכלול הגורמים המעורבים באירועי דריסות‬
‫של בעלי חיים‪ .‬מתוך כך המחקר מציע מודל חדשני מבוסס על מנגנונים אקולוגיים‪-‬ביולוגיים‪ ,‬באמצעותו ניתן‬
‫לספק המלצות מושכלות להפחתת שיעור הדריסות הן בכבישים קיימים‪ ,‬והן בבניית כבישים חדשים‪.‬‬
‫אזור המחקר‪:‬‬
‫שפלת יהודה וכבישי‬
‫המחקר‬
‫מיפוי שימושי הקרקע‬
‫באזור המחקר‬
‫‪85‬‬
‫תמונות מייצגות של אירועי דריסה מאזור המחקר‪:‬‬
‫תוצאות‪:‬‬
‫במהלך המחקר נאספו נתונים על ‪ 808‬אירועי דריסה של בע"ח שונים‪ .‬מתוך נתונים אלו‪ ,‬בידדתי ‪ 022‬אירועי‬
‫התנגשות של רכב עם בע"ח אשר הציעו כי היוו סכנה לחיי אדם‪ .‬המינים אשר נכללו בניתוח זה (תאונות‬
‫מסכנות חיי אדם) מוצגים בטבלה ‪ .0‬מינים אלו נלקחו בחשבון כמסכנים חיי אדם כאשר הם מעורבים בתאונות‬
‫‪86‬‬
‫עם רכבים על סמך משקל וגודל גוף ממוצע ואופי הפגיעה (כלומר‪ :‬זוחלים‪ ,‬מלבד צב יבשה‪ ,‬לא נלקחו בחשבון‬
‫מאחר והם על פני הקרקע‪ ,‬אך עופות כן נלקחו בחשבון בעקבות פוטנציאל התנגשות בשמשה)‪.‬‬
‫טבלה ‪ :1‬מינים אשר נכללו בניתוח המרחבי של תאונות מסכנות חיי אדם‪.‬‬
‫מין‬
‫מספר תצפיות‬
‫אנפית בקר‬
‫‪1‬‬
‫עורב‬
‫‪6‬‬
‫עורב אפור‬
‫‪2‬‬
‫עקב עיטי‬
‫‪1‬‬
‫תנשמת‬
‫‪18‬‬
‫תרנגול‬
‫‪11‬‬
‫ציפור גדולה ב"מ‬
‫‪1‬‬
‫צב יבשה‬
‫‪5‬‬
‫דורבן‬
‫‪4‬‬
‫חתול בית‬
‫‪49‬‬
‫נמייה‬
‫‪53‬‬
‫צבי‬
‫‪1‬‬
‫שועל‬
‫‪39‬‬
‫תן‬
‫‪41‬‬
‫גירית מצויה‬
‫‪1‬‬
‫סה"כ‬
‫‪322‬‬
‫על מנת לזהות את המאפיינים אשר משפיעים על מספר הדריסות במקטע כביש מסוים‪ ,‬ביצעתי ניתוח של‬
‫השפעת מאפיינים מרחביים בסקאלות שונות על מספר הדריסות הנצפה במקטעי כביש באורך של ‪ 011‬מטר‪.‬‬
‫כלומר‪ ,‬מקטעי הכביש אופיינו על ידי משתנים מרחביים בסקאלות (טווחים מרחביים) שונות ומודלים‬
‫סטטיסטיים שונים נבדקו לבירור ההשפעה של כל אחד מהמשתנים על הסתברות אירועי דריסה במקטעי כביש‬
‫באורך של ‪ 011‬מטר‪.‬‬
‫ראשית‪ ,‬ביצעתי ניתוח של השפעת מאפיינים בסקאלה המקומית על מספר אירועי הדריסה‪ .‬מניתוח זה מצאתי‬
‫כי מקטעי כביש מוארים מראים מספר גבוה יותר של דריסות לעומת מקטעי כביש ללא תאורה (‪.)p<<0.01‬‬
‫בנוסף‪ ,‬מצאתי שמקטעי כביש שנמצאים באותו הגובה של סביבתם הקרובה‪ ,‬גם כן מראים מספר גבוה יותר של‬
‫דריסות לעומת מקטעי כביש מוגבהים‪/‬שקועים (‪ .)p<<0.01‬לבסוף‪ ,‬מצאתי שהמרחק למעבר בע"ח הקרוב‬
‫‪87‬‬
‫ביותר גם כן משפיע על מספר הדריסות במקטע כביש מסוים (‪ .)p=0.03‬המרחק ממעבר בע"ח הקרוב ביותר‬
‫נמצא במתאם שלילי עם מספר הדריסות‪ .‬כלומר‪ ,‬יותר דריסות נמצאו בקרבת מעברים מאשר במרחק רב מהם‪.‬‬
‫לאחר מכן‪ ,‬בדקתי את השפעת ההקשר המרחבי בסקאלת הנוף על מספר אירועי הדריסה במקטעי הכביש‬
‫השונים‪ .‬תחילה‪ ,‬חישבתי את אחוז הכיסוי של שימושי קרקע שונים ברדיוס של ‪ 0,111‬מטר סביב כל מקטע‬
‫כביש‪ .‬לאחר מכן‪ ,‬ביצעתי הליך של ברירת מודלים וחישוב החשיבות היחסית של כל משתנה לרמת הדיוק של‬
‫תחזית המודל‪ .‬ניתוח זה הראה שההשפעה הגדולה ביותר הינה של יער טבעי (אינדקס חשיבות = ‪.)1.98‬‬
‫השפעה זו נמצאה במתאם שלילי בין אחוז הכיסוי של יער טבעי ברדיוס של ‪ 0,111‬מטר ומספר הדריסות‬
‫(‪ .) p<<0.01‬כלומר‪ ,‬ככל שאחוז הכיסוי של יער טבעי מסביב למקטע כביש גבוה יותר‪ ,‬מספר הדריסות צפוי‬
‫להיות נמוך יותר‪ .‬בנוסף‪ ,‬נמצאה השפעה בינונית לאחוז הכיסוי של יער נטוע מעשה אדם וישובים (אינדקס‬
‫חשיבות = ‪ 1.28‬ו‪ ,1.28 -‬בהתאמה)‪ .‬השפעות אלו נמצאו במתאם חיובי עם מספר הדריסות (‪ p=0.03‬לשני‬
‫שימושי הקרקע)‪ .‬כלומר‪ ,‬ככל שאחוז כיסוי השטח של שימושי קרקע אלו גבוה יותר‪ ,‬מספר הדריסות צפוי‬
‫להיות גבוה יותר‪.‬‬
‫לבסוף‪ ,‬בדקתי השפעה משותפת של משתנים משתי הסקאלות (המקומית והנוף) על מספר הדריסות במקטע‬
‫כביש‪ .‬מניתוח זה מצאתי כי החשיבות הגבוהה ביותר הינה של שלושה משתנים‪ :‬אחוז כיסוי של יער נטוע‬
‫מעשה אדם‪ ,‬תאורה וחיבור הכביש עם פני הקרקע (אינדקס חשיבות = ‪ 1.99‬לשלושת המשתנים)‪ .‬בנוסף‪,‬‬
‫מצאתי חשיבות מסוימת גם לאחוז כיסוי קרקע של יער טבעי (אינדקס חשיבות = ‪ .)1..0‬תוצאות אלו מדגישות‬
‫את החשיבות של כלל ההקשר המרחבי (שתי הסקאלות יחדיו) של מקטעי כביש בחיזוי מספר דריסות‪.‬‬
‫מסקנות העבודה ובניית מודל יישומי פוטנציאלי‪:‬‬
‫בהתאם לתוצאות מחקר זה‪ ,‬ניתן לסכם כי ישנם מספר מאפיינים אשר משפיעים על ההסתברות לתאונות עם‬
‫בע"ח המסכנות חיי אדם‪ .‬מקטעי כביש אשר הינם בעלי סיכון גבוה יותר לתאונות אלו הינם‪:‬‬
‫‪‬‬
‫מוארים‬
‫‪‬‬
‫נמצאים בגובה פני הסביבה‬
‫‪‬‬
‫נמצאים בסביבת יער נטוע מעשה אדם‬
‫‪‬‬
‫נמצאים בקרבה ליישובים‬
‫מחקר זה שאף להבנת המנגנונים הפועלים בסקאלות שונות ואשר מעורבים ביצירת דגמי תאונות‪ .‬זאת‪ ,‬על מנת‬
‫לאתר משתנים מסבירים בעלי השפעה על מנגנונים אלו‪ ,‬ולמעשה על מיקום ואופי חציית הכביש ע"י בע"ח‪.‬‬
‫בעקבות כך‪ ,‬התאפשרה בניית מודל מרחבי תלוי סקאלה לחיזוי אתרים בעלי סיכון גבוה לתאונות‪ .‬יישום עתידי‬
‫של מודל זה יכול לשמש למתן המלצות ודרכי פעולה על מנת להגדיל את היעילות של תוכניות פעולה ואמצעים‬
‫להפחתת תאונות דרכים בהן מעורבים בע"ח‪ .‬בכדי להבין את מורכבות המודל‪ ,‬נתאר מצב פשטני של תנועת‬
‫‪88‬‬
‫בע"ח במרחב ‪ --‬אם במהלך תנועה על מנת למלא אחר צרכיו הקיומיים‪ ,‬נתקל בעל החיים בכביש‪ ,‬עליו להחליט‬
‫אם לחצותו‪ .‬לכן‪ ,‬שטף התנועה של בע"ח במקטע כביש מסוים נקבע על פי הקשרו המרחבי בסקאלה רחבה‬
‫(סקאלת הנוף)‪ .‬לאחר מכן‪ ,‬בנקודת החצייה עצמה‪ ,‬על בע"ח לבחור בין שתי אפשרויות‪ :‬לחצות על גבי הכביש‬
‫או לחפש ולעשות שימוש במבנה חציה (אם קיים במקטע כביש זה)‪ .‬במקרה ובעל החיים חצה על גבי הכביש‬
‫ייתכן ויחצה בהצלחה וייתכן שיידרס‪.‬‬
‫מטריצת החלטות זו מוצגת באיור הבא המהווה את הבסיס למודל המוצע‪:‬‬
‫תרשי של השו ות המרחבית‬
‫שימושי קרקע‪ ,‬פיזור‬
‫המשאבי ‪ .‬ו ר על ב י‬
‫שכבת מידע ממוחשבת‬
‫השפעה על רמות פעילות של בע ח במרחב‬
‫ע י יתוב ת ועה למק עי כביש שו י‬
‫מפה תיאורית של הד‬
‫ה ו ר של רמות פעילות‬
‫במרחב‬
‫מאפיי י מקומיי‬
‫משפיעי על אופי‬
‫ח יית הכביש‬
‫ותו אותיה‬
‫כאשר‪ Ax ,‬מייצג את שטף התנועה במקטע כביש מסוים‪ .‬שטף זה הינו תוצאה של ההקשר המרחבי בסקאלה‬
‫רחבה (סקאלת הנוף) של אותו מקטע כביש‪ p(Surface) .‬היא ההסתברות שבתוך מקטע כביש בע"ח יחצה על‬
‫גבי הכביש‪ .‬בהתאמה‪ p(Structure) ,‬היא ההסתברות המשלימה‪ ,‬קרי שבע"ח יחצה במבנה‪ p(RK) .‬היא‬
‫ההסתברות שבע"ח שחצה על גבי הכ ביש יידרס‪ .‬המרכיבים ההסתברותיים הללו מושפעים כולם מתהליכים‬
‫בסקאלה הצרה (הסקאלה המקומית)‪ ,‬ולכן מהמאפיינים המקומיים של כל מקטע כביש‪.‬‬
‫כאמור‪ ,‬כמקובל כיום בחקר של דגמי תנועה במרחב‪ ,‬המחקר הנוכחי נסמך על גישה תלוית סקאלה‪ .‬על בסיס‬
‫גישה זו‪ ,‬מוצע בזאת מודל המבוסס על כך שבכל סקאלה פועלים תהליכים שונים‪ ,‬אשר תורמים בצורה שונה‬
‫לדגם המתקבל‪ .‬בסקאלת הנוף‪ ,‬תהליכים המשפיעים על רמת הפעילות של בע"ח במקטע מסוים ייקבעו את‬
‫שטף התנועה בו‪ .‬בסקאלה המקומית‪ ,‬מתקיימים תהליכים המשפיעים על סבירות החצייה במבנה‪ ,‬כמו גם‪ ,‬על‬
‫הסבירות לתאונ ת דרכים במקרה של חצייה על גבי הכביש‪ .‬כל אלו באים באינטראקציה עם דחף התנועה‬
‫(‪ )vagility, Carr & Fahrig 2001‬של המין הספציפי ותפיסת הסיכון של הפרט בשולי הכביש‪ .‬תהליכים אלו‬
‫‪89‬‬
‫קובעים את אופי החצייה ותוצאותיה‪ .‬בנוסף‪ ,‬המחקר האקולוגי העכשווי מציע כי קיימת אינטראקציה בין‬
‫התהליכים הפועלים בסקאלות השונות‪ .‬לפיכך‪ ,‬כאשר באים להבין את תנועתם של בע"ח במרחב‪ ,‬כמו גם את‬
‫אזורי החציה והסבירות לתאונות‪ ,‬יש לקחת בחשבון את התהליכים האקולוגיים המרחביים בסקאלות שונות‬
‫ואת הפסיפס הנופי של שימושי קרקע אנטרופוגניים‪.‬‬
‫מסקנות אלו והמודל המוצע שנובע מהן‪ ,‬עומדים בקנה אחד עם עבודה שבוצעה ע"י ד"ר עמית דולב בגליל‬
‫(דולב ‪ .)011.‬עבודה זו הראתה ששיעור הדריסות לק"מ של שועלים ותנים בכביש ראשי (המאופיין בנתיב‬
‫אחד לכל כיוון) היה גבוה פי ארבע משיעור הדריסות בכביש מהיר (עם גדר הפרדה מסוג ניו ג'רזי)‪ .‬לפיכך‪,‬‬
‫נראה כי מספר התאונות בכבישים בעלי עצימות בינונית (מדד המשקלל מספר פרמטרים כגון רוחב הכביש‪,‬‬
‫מהירות נסיעה‪ ,‬נפח תנועה ונוכחות מחסום ג'רזי) צפוי להיות גבוה יותר מאשר בכבישים מהירים‪ .‬כבישים אלו‬
‫הינם בעיקר כבישים בין עירוניים שסלולים בגובה פני הסביבה‪ .‬בכבישים אלו‪ ,‬מהירות הנסיעה גבוהה ולכן‬
‫הסיכון לפגיעה ע"י כלי‪-‬רכב הוא גבוה‪ .‬בנוסף‪ ,‬החיבור הפיזי של הכביש עם הסביבה מאפשר נגישות לתנועת‬
‫בע"ח ומעלה את שטף התנועה של בע"ח על גבי הכביש‪ .‬בעקבות כך‪ ,‬גם הסבירות לתאונות דרכים הינה גבוהה‬
‫יותר בכבישים מסוג זה (שקדי ושדות ‪ .)0112‬ראוי לציין‪ ,‬כי ארץ ישראל ייחודית‪ ,‬בכך שמרבית הכבישים‬
‫הסלולים בה הינם בעלי אופי זה‪ .‬נכון לשנת ‪ ,0118‬בישראל ‪ 08,911‬ק"מ כביש סלול‪ ,‬מתוכם רק ‪ 021‬ק"מ‬
‫מוגדרים ככביש מהיר‪ .‬לעומת זאת‪ ,‬כבישים בעלי עצימות בינונית מהווים כ‪ .0% -‬מסך השטח של כבישים‬
‫סלולים בארץ (עפ"י נתונים רשמיים מאתר משרד התחבורה)‪.‬‬
‫המלצות‪:‬‬
‫ההמלצות המובאות להלן הן בהתאם לתוצאות מחקר זה ולסקר הספרות שבוצע כחלק מהמחקר‪.‬‬
‫‪‬‬
‫גופי שמירת הטבע מחד ומשרדי התשתיות והעוסקים בדבר מאידך ‪,‬חייבים להסתכל על סוגיית ממשק‬
‫כביש‪-‬סביבה בראיה מרחבית גדולה יותר‪ .‬כל תוכניות הפעולה הנוגעות לדריסות ומעברי בע"ח‬
‫חייבות לקחת בחשבון את הצורך ורצון בע"ח לנוע ממקום למקום בסקאלות גדולות;‬
‫‪‬‬
‫באפיון ותכנון של תוכניות להפחתת שיעור תאונות המערבות בעלי חיים‪ ,‬יש להגדיר מטרות ברורות‪.‬‬
‫מטרות כמותיות‪ ,‬הניתנות לבדיקה‪ ,‬הינן עדיפות על מטרות איכותיות;‬
‫‪‬‬
‫יש להתמקד בכבישים ראשיים בעלי עצימות בינונית‪ .‬כבישים אלו הינם בעיקר כבישים בין עירוניים‬
‫שסלולים בגובה פני הסביבה‪ .‬בכבישים אלו‪ ,‬מהירות הנסיעה גבוהה ולכן הסיכון לפגיעה ע"י כלי‪-‬‬
‫רכב הוא גבוה‪ .‬בנוסף‪ ,‬החיבור הפיזי של הכביש עם הסביבה מאפשר נגישות לתנועת בע"ח ובכך‬
‫תורם להגברת שטף התנועה של בע"ח על גבי הכביש‪ .‬בעקבות כך‪ ,‬גם הסבירות לתאונות דרכים‬
‫הינה גבוהה יותר;‬
‫‪‬‬
‫יש להתמקד במקטעי כביש מוארים;‬
‫‪90‬‬
‫‪‬‬
‫מומלץ להתמקד במקטעי כביש אשר נמצאים בגובה פני הסביבה;‬
‫‪‬‬
‫מומלץ להתמקד במקטעי כביש אשר נמצאים בקרבה ליער נטוע מעשה אדם ובקרבת יישובים;‬
‫‪‬‬
‫רצוי להציב שלטי אזהרה במקטעי כביש אלו‪ .‬מאחר ושלטים אלו נוטים לאבד מיעילותם במהלך הזמן‪,‬‬
‫מומלץ להציבם בתקופות המסוכנות יותר לתאונות עם בע"ח‪ ,‬קרי בתקופת האביב והקיץ‪ .‬כמו כן‪,‬‬
‫מומלץ לשנות את אופי שלטים אלו מעת לעת ולשנות במעט את מיקומם;‬
‫‪‬‬
‫יש לבנות מעברים ייעודים לבע"ח במקטעי כביש אלו ולהציב הכוונה (כמו‪ ,‬גידור) לבע"ח למבנים‬
‫אלו‪ .‬כלומר‪ ,‬רצוי לנתב את תנועת בע"ח למבנים אלו בעזרת גידור של כ‪ 01 -‬מטר מכל אחד מצדי‬
‫המעבר;‬
‫‪‬‬
‫יש לדאוג כי המעברים‪ ,‬כמו גם הגידור לצדי הכביש‪ ,‬יהיו מתוחזקים באופן שוטף‪ .‬כלומר‪ ,‬ניקוז‪,‬‬
‫ניקוי‪ ,‬תיקון קירות ופלטפורמת המעבר‪ ,‬גיזום וכל פעילות אחרת בהתאם לצורך הפרטני של המעבר‬
‫הספציפי;‬
‫‪‬‬
‫יש לדאוג לניטור יעילות המעברים והתאמתם בפועל למינים ספציפיים בהתאם לדרישות המין;‬
‫‪‬‬
‫יש לדאוג לניטור מספר הדריסות ו‪/‬או התאונות עם בע"ח במקטעי הכביש;‬
‫‪‬‬
‫יש לוודא רציפות מעבר על פני הכביש לבע"ח אשר נקלע לאזור הכביש‪ .‬בע"ח עלולים להיקלע‬
‫לכבישים סואנים במידה וגידור חסר‪ ,‬לא יעיל‪ ,‬או לא מתוחזק היטב‪ .‬כאשר בע"ח נקלע לאזור הכביש‬
‫יש להבטיח את יכולתו לעבור לצד הכביש‪ .‬יצירת חיץ מבודד בין נתיבי הכביש עלולה לגרום נזק‬
‫עצום לבע"ח על ידי הגדלת סיכוייהם להידרס‪ ,‬כמו גם על ידי מניעת רציפות בית הגידול משני צדי‬
‫הכביש במקרים שמעבר על פני הכביש עשוי להיות יעיל (למשל בכבישים קטנים בעלי תנועת מכוניות‬
‫דלה)‪ .‬מחסומי בטון מסוג "ניו ג'רזי" גורמים לנזק רב לבע"ח מהסיבות שצוינו לעיל‪ .‬איי תנועה עם‬
‫צמחייה טבעית עשויים לעזור לבע"ח לעצור במרכז התוואי ולהקטין בכך את סיכויי התאונה‪ .‬אולם‪,‬‬
‫יש לבחון פתרון זה לגופו מאחר והתברר שבמקומות מסוימים בעולם איי תנועה אלו היוו מקור משיכה‬
‫למינים שונים‪ ,‬וכתוצאה מכך עלו סיכויי הידרסותם‪.‬‬
‫תרומת המחקר למטרות "קרן רן נאור לבטיחות בדרכים"‬
‫המחקר המתואר להלן‪ ,‬עומד בקנה אחד עם היוזמה לקדם את הבטיחות בדרכים ולהפחית את שיעורן של תאונות‬
‫מסכנות חיים‪ .‬כל זאת‪ ,‬ע"י ביסוס והעשרת הידע המשמש לגיבוש מדיניות הרשות הלאומית לבטיחות בדרכים‪,‬‬
‫וככלי עזר לקבלת החלטות על פעולות הרשות‪ .‬בנוסף‪ ,‬מחקר זה מאפשר להעריך את הפעילות המתבצעת‬
‫בתחום זה כיום‪.‬‬
‫ראוי לציין‪ ,‬כי המעבדה לאקולוגיה מרחבית באוניברסיטת בן גוריון שבנגב הינה מערכת המחקר המובילה בארץ‬
‫העוסקת במחקר מדעי בתחום זה‪ ,‬וכי‪ ,‬עד כמה שידוע לנו‪ ,‬זהו המחקר המקיף ביותר שבוצע בארץ בנושא‬
‫הפחתת הסיכון לחיי אדם הנובע מיחסי הגומלין בין כבישים לסביבה הטבעית‪.‬‬
‫‪91‬‬
.‫אנו מודים לקרן רן נאור לבטיחות בדרכים על העזרה והתמיכה לקיום המחקר‬
‫רשימת המקורות בהם נעשה שימוש לכתיבת הנספח‬
Alexander, S. M., N. M. Waters, and P. C. Paquet. 2005. Traffic volume and highway permeability
for a mammalian community in the Canadian Rocky Mountains. Canadian GeographerGeographe Canadien 49:321-331.
Andrews, K. M., and J. W. Gibbons. 2005. How do highways influence snake movement?
Behavioral responses to roads and vehicles. Copeia:772-782.
Baker, P. J., C. V. Dowding, S. E. Molony, P. C. L. White, and S. Harris. 2007. Activity patterns of
urban red foxes (Vulpes vulpes) reduce the risk of traffic-induced mortality. Behavioral
Ecology 18:716-724.
Bellis, M. A., S. D. Jackson, C. R. Griffin, P. S. Warren, and A. O. and Thompson. 2007. Utilizing a
Multi-Technique, Multi-Taxa Approach to Monitoring Wildlife Passageways on the
Bennington Bypass in Southern Vermont. eScholarship, UC Davis: Road Ecology Center.
Burnham, Kenneth P., and David R. Anderson. 2002. Model Selection and Multimodel Inference:
A Practical Information-Theoretic Approach. Second ed. New York: Springer-Verlag.
Carr, L. W., and L. Fahrig. 2001. Effect of road traffic on two amphibian species of differing vagility.
Conservation Biology 15:1071-1078.
Conover, M. R., W. C. Pitt, K. K. Kessler, T. J. Dubow, and W. A. Sanborn. 1995. Review of Human
Injuries, Illnesses, and Economic-Losses Caused by Wildlife in the United-States. Wildlife
Society Bulletin 23:407-414.
Forman, R. T. T., and L. E. Alexander. 1998. Roads and their major ecological effects. Annual
Review of Ecology and Systematics 29:207.+Glista, D. J., T. L. DeVault, and J. A. DeWoody. 2009. A review of mitigation measures for reducing
wildlife mortality on roadways. Landscape and Urban Planning 91:1-7.
Grilo, C., J. A. Bissonette, and M. Santos-Reis. 2009. Spatial-temporal patterns in Mediterranean
carnivore road casualties: Consequences for mitigation. Biological Conservation 142:301313.
Inbar, M., U. Shanas, and I. Izhaki. 2002. Characterizing of road accidents in Israel involving large
mammals. Israel Journal of Zoology 48:10.
Joyce, T. L., and S. P. Mahoney. 2001. Spatial and temporal distributions of moose-vehicle
collisions in Newfoundland. Wildlife Society Bulletin 29:281-291.
Laurance, W. F., M. Goosem, and S. G. W. Laurance. 2009. Impacts of roads and linear clearings
on tropical forests. Trends in Ecology & Evolution 24:659-669.
Malo, J. E., F. Suarez, and A. Diez. 2004. Can we mitigate animal-vehicle accidents using predictive
models? Journal of Applied Ecology 41:701-710.
Mata, C ,.I. Hervas, J. Herranz, J. E. Malo, and F. Suarez. 2009. Seasonal changes in wildlife use of
motorway crossing structures and their implication for monitoring programmes.
Transportation Research Part D-Transport and Environment 14:447-452.
Mazerolle, M .J., M. Huot, and M. Gravel. 2005. Behavior of amphibians on the road in response
to car traffic. Herpetologica 61:380-388.
Putman, R. J. 1997. Deer and road traffic accidents: Options for management. Journal of
Environmental Management 51:43-57.
Turner, M .G. 2005. Landscape ecology: What is the state of the science? Annual Review of
Ecology Evolution and Systematics 36:319-344.
92
‫‪van der Ree, R., J. A. G. Jaeger, E. A. van der Grift, and A. P. Clevenger. 2011. Effects of Roads and‬‬
‫‪Traffic on Wildlife Populations and Landscape Function: Road Ecology is Moving toward‬‬
‫‪Larger Scales. Ecology and Society 16.‬‬
‫‪Venables, W. N., D. M. Smith and the R Development Core Team. 2011. An Introduction to R.‬‬
‫‪Notes on R: A Programming Environment for Data Analysis and Graphics Version 2.13.2‬‬
‫)‪(2011-09-30‬‬
‫‪Vuilleumier, S., and R. Metzger. 2006. Animal dispersal modelling: Handling landscape features‬‬
‫‪and related animal choices. Ecological Modelling 190:159-170.‬‬
‫בקי‪ ,‬א‪ . 2002 .‬מעברי בעלי חיים בכבישים‪ .‬מסמך רקע לקביעת מדיניות‪ .‬החברה לזואולוגיה בישראל‬
‫ורשות הטבע והגנים‪.‬‬
‫גוטמן‪ ,‬ר‪ ,.‬י‪ .‬סיני‪ ,‬א‪ .‬שדות‪ ,‬י‪ .‬שקדי‪ .2002 .‬השפעה של התנועה בכבישי ישראל על התמותה של בעלי‬
‫חיים‪ ,‬ובחינת יעילות מעברי בעלי החיים הקיימים‪ .‬רשות הטבע והגנים‪.‬‬
‫דולב‪ ,‬ע‪ .2002 .‬מודל התפשטות מרחבי של מחלת הכלבת המבוסס על דינאמיקה של אוכלוסיות‬
‫השועל המצוי בגליל‪ ,‬ככלי לפיתוח דגמי פיזור אופטימאליים של פיתיונות חיסון כלבת‪.‬‬
‫עבודת תזה במסגרת הדרישות לקבלת תואר "דוקטור לפילוסופיה"‪.‬‬
‫שקדי‪ ,‬י‪ ,.‬א‪ .‬שדות‪ .2002 .‬מעבר בעלי חיים בכבישים‪ .‬מדיניות והמלצות לפעולה‪ .‬פרסומי חטיבת‬
‫המדע‪ .‬רשות הטבע והגנים‪.‬‬
‫‪93‬‬
‫מהמעבר הקרוב ביותר וגובה הכביש ביחס לפני הסביבה ‪ --‬משפעים בברור על שיעור הדריסות‪ .‬באופן דומה‪ ,‬גורמים‬
‫המאפיינים את המעברים בסקאלה המרחבית המקומית ‪ --‬כדוגמת‪ :‬כיסוי צומח ורמת שימוש ע"י בני אדם ‪ --‬משפיעים‬
‫על דגמי השימוש במעברים‪ .‬בסקאלה רחבה יותר‪ ,‬מצאתי שגורמים המתארים את ההקשר המרחבי של מקטעי כביש‬
‫ומעברים בסקאלה נופית‪ ,‬לדוגמת פסיפס שימושי הקרקע‪ ,‬משפיעים בברור על תבנית החדירות של הכביש (מה‬
‫שמתבטא בשיעור דריסות ושימוש במעברים)‪ .‬בהתאם‪ ,‬אני טוען שפסיפס שימושי הקרקע מבטא למעשה את חלוקת‬
‫המשאבים‪ ,‬ובהתאם מכתיב את שכיחותם של מינים מסויימים‪ .‬כמו כן‪ ,‬פסיפס זה מכתיב את נתיבי התנועה העיקריים‬
‫של בע"ח‪ ,‬מאחר ושימושי קרקע שונים מהווים מוקדי משיכה או דחייה לתנועת בע"ח‪.‬‬
‫לסיכום‪ ,‬תוצאות המחקר הנוכחי מצביעות על כך שתבניות של 'חדירות כבישים' מעידות על שני תהליכים עיקריים‪:‬‬
‫‪ .1‬ברמת הסקאלה הנופית‪ ,‬פעילותם של חיות בר (שטף התנועה) נקבע ע"י הנגישות למקטע כביש מסוים וההקשר‬
‫המרחבי שלו‪.‬‬
‫‪ .2‬ברמת הסקאלה המקומית‪ ,‬מאפיינים מרחביים משפיעים על ההחלטה של כל פרט בנוגע לאפשרויות השונות‬
‫לחציית הכביש‪ ,‬ולפיכך‪ ,‬משפיעים בסיכומו של דבר על אירועי חצייה בתוך אותו מקטע כביש‪.‬‬
‫בהתאם לכך‪ ,‬תהליכים ומשתנים אלו צריכים להילקח בחשבון בעת אפיון ותכנון של תוכניות הקלה להפחתת השפעת‬
‫הקיטוע ההרסנית של כבישים‪.‬‬
‫ב‬
‫תקציר‬
‫כבישים ותנועת רכבים מייצרים קיטוע מרחבי ומהווים מחסום לתנועת בעלי חיים בטבע‪ .‬מחקרים מראים שקיטוע נוף‬
‫ע"י כבישים מהווה סיכון לרמות טקסונומיות שונות‪ :‬חסרי חוליות‪ ,‬דו‪-‬חיים‪ ,‬זוחלים‪ ,‬ציפורים ויונקים‪ .‬השפעה‬
‫שלילית ומזיקה זו טומנת בחובה עלות אקולוגית משמעותית הכוללת אובדן בתי גידול‪ ,‬עליה בבידוד התוך והבין מיני‪,‬‬
‫תמותה של חיות בר‪ ,‬הפחתת הגישה למשאבים בסיסיים והפרעה לתהליכים אקולוגיים שונים‪ .‬בנוסף‪ ,‬מטבע הדברים‪,‬‬
‫כבישים תורמים לעליה בשיעור תאונות הדרכים בין חיות‪-‬בר לרכבים‪ ,‬גורם שמהווה את אחת ההשפעות השליליות‬
‫המוחשיות ביותר של האנושות על בעלי חיים בטבע‪ .‬כמות גוברת של ספרות מקצועית בתחום ה‪Road Ecology -‬‬
‫גורסת כי קיטוע נוף ע"י כבישים עלול להוות גורם מרכזי בתמותת בעלי חוליות‪ ,‬ולפיכך להוות חסם לגידול אוכלוסיית‬
‫הבר‪ .‬יתר על כן‪ ,‬מאמרים עדכניים מראים כי כבישים מהווים את הגורם מספר אחת לתמותת חיות בר‪.‬‬
‫מכיוון שתנועת הרכבים מהווה למעשה יותר מסננת לתנועת חיות מאשר מחסום פיזי מוחלט‪ ,‬האפקט שנוצר זוכה‬
‫לעיתים לכינוי‪( Relative Road Permeability :‬חדירות כביש יחסית)‪ .‬ככל שתנועת בעלי החיים בחציית הכביש גבוהה‬
‫יותר‪ ,‬כך עולה שיעור החדירות היחסית של הכביש‪ .‬בספרות המקצועית ידוע כי 'חדירות הכביש' יכולה להשפיע ולשנות‬
‫את הרכב חברות בעלי החיים‪ ,‬ליצור חברות ו‪/‬או אוכלוסיות מבודדות‪ ,‬להוריד את ההטרוגניות והשונות הביולוגית‬
‫ולהעלות את הסיכון להכחדת מינים‪ .‬בנוסף‪' ,‬חדירות הכביש' עלולה להוות גורם משפיע על התנהגות בעלי חיים‪.‬‬
‫מודלים ככלל‪ ,‬וכאלו המשלבים התייחסות לסקאלות מרחביות בפרט‪ ,‬אשר מאפשרים חיזוי של תבניות המייצגות את‬
‫חדירותם של כבישים‪ ,‬צפויים לתרום משמעותית להפחתת ההשפעה השלילית עבור האדם וחיות הבר כאחד‪ .‬מטרתו‬
‫המרכזית של מחקר זה‪ ,‬הינה לזהות משתנים )‪ (determinants‬המצויים במתאם עם תבניות של 'חדירות כבישים'‬
‫(לדוגמת דריסות ושימוש במבנים המשמשים למעבר בע"ח) בסקאלות מרחביות משתנות‪.‬‬
‫במסגרת המחקר‪ ,‬זיהיתי משתנים מסבירים עבור דריסות (‪ )road-kill‬ודגמי שימוש במעברי כביש בתצורות שונות‬
‫(‪ )Passage-use Patterns‬ע"י השוואה של מגוון מודלים בעזרת התוכנה הסטטיסטית '‪ 'R‬וע"י מתודולוגיית ברירת‬
‫מודלים (‪ .)Model Selection‬המודלים מבוססים על מנגנונים בעלי אופי ביולוגי‪-‬אקולוגי ומורכבים ממספר משתנים‬
‫המסבירים את השונות הנופית‪ ,‬תכונות הכביש ומאפייני המעבר‪.‬‬
‫המחקר כלל סקירה של שלושה כבישים הממוקמים בצפון שפלת יהודה בישראל (סה"כ אורך הכבישים הינו כ‪ 24-‬ק"מ)‪.‬‬
‫סקרי דריסות בוצעו ע"י נסיעה איטית עם שחר‪ 4-10 ,‬ימים בכל חודש‪ .‬דפוסי שימוש במעברים (שיעור חצייה במעבר‬
‫ולחץ טריפה) תועדו באמצעות מצלמות אינפרא‪-‬אדום וניסויי טריפה (ע"י הנחה מלאכותית של קני ציפורים; ‪Artificial‬‬
‫‪ .)Nest Predation Experiment‬השונות הנופית נותחה והוטמעה בתוכנות ‪ ,GIS‬באמצעות שכבת אורתופוטו (תצלומי‬
‫אויר מיושרים) ובאמצעות מדידות ידניות בשטחי הניסוי‪ .‬בנוסף‪ ,‬ביצעתי ניתוח נתונים בעזרת תוכנות גיאו‪-‬‬
‫סטטיסטיות וכלי אנאליזה לסטטיסטיקה רב‪-‬משתנית‪ ,‬במטרה לאתר מתאמים בין משתני הכביש‪ ,‬ההקשר המרחבי‬
‫שלהם ותכונות מין ספציפי‪ ,‬לבין תצפיות של דריסות בעלי חיים ודגמים של שימוש במעברים‪.‬‬
‫בסל הכל‪ ,‬תועדו ‪ 817‬דריסות‪ ,‬כאשר מספר הדריסות הגבוה ביותר נמצא בכביש הסואן ביותר‪ .‬שימוש בניתוח‬
‫סטטיסטי מבוסס על פונקציית ‪ ,Ripley's k‬המשמשת לניתוח דגמים מרחביים‪ ,‬הראה שמיקומן הפיזי של הדריסות‬
‫נמצא בהתקבצויות מרחביות בכל שלושת הכבישים‪ .‬עם זאת‪ ,‬הרזולוציה המרחבית שבה מתקיימים אותם מצבורים‬
‫של דריסות‪ ,‬משתנה בין הכביש הסואן ביותר לשני הכבישים הסואנים פחות‪ ,‬תוצאה שעשויה להצביע על כך שהכביש‬
‫הסואן ביותר מהווה גורם משמעותי יותר לקיטוע נוף‪ .‬בנוסף‪ ,‬מצאתי השפעה משמעותית של ההקשר המרחבי של‬
‫מקטעי כביש מסויימים ומעברים‪ ,‬עם גורמים מסקאלות שונות המשפיעים על שיעור הדריסות ודגמי החצייה במעברים‪.‬‬
‫מצאתי שמספר גורמים המתארים את ההקשר המרחבי של מקטעי כביש בסקאלה המקומית ‪ --‬כדוגמת‪ :‬תאורה‪ ,‬מרחק‬
‫א‬
‫הצהרת תלמיד המחקר עם הגשת עבודת הדוקטור לשיפוט‬
‫אני החתום מטה מצהיר‪/‬ה בזאת‪( :‬אנא סמן)‪:‬‬
‫___ חיברתי את חיבורי בעצמי‪ ,‬להוציא עזרת ההדרכה שקיבלתי מאת מנחה‪/‬ים‪.‬‬
‫___ החומר המדעי הנכלל בעבודה זו הינו פרי מחקרי מתקופת היותי תלמיד‪/‬ת מחקר‪.‬‬
‫___ בעבודה נכלל חומר מחקרי שהוא פרי שיתוף עם אחרים‪ ,‬למעט עזרה טכנית‬
‫הנהוגה בעבודה ניסיונית‪ .‬לפי כך מצורפת בזאת הצהרה על תרומתי ותרומת שותפי‬
‫למחקר‪ ,‬שאושרה על ידם ומוגשת בהסכמתם‪.‬‬
‫תאריך ________ שם התלמיד‪/‬ה __יואב אבניאון__ חתימה ___________‬
‫העבודה נעשתה בהדרכת‬
‫פרופ' ירון זיו‬
‫המחלקה למדעי החיים‪ ,‬הפקולטה למדעי הטבע‬
‫אוניברסיטת בן גוריון בנגב‬
‫ניתוח מרחבי של חדירות כבישים לתנועת בע"ח‪:‬‬
‫מדגמים לתהליכים‬
‫מחקר לשם מילוי חלקי של הדרישות לקבלת תואר "דוקטור לפילוסופיה"‬
‫מאת‬
‫יואב אבניאון‬
‫הוגש לסינאט אוניברסיטת בן גוריון בנגב‬
‫אישור המנחה ____________________‬
‫אישור דיקן בית הספר ללימודי מחקר מתקדמים ע"ש קרייטמן ___________‬
‫‪ 81‬מרץ ‪3182‬‬
‫ט"ז אדר ב' תשע"ד‬
‫באר שבע‬
‫ניתוח מרחבי של חדירות כבישים לתנועת בע"ח‪:‬‬
‫מדגמים לתהליכים‬
‫מחקר לשם מילוי חלקי של הדרישות לקבלת תואר "דוקטור לפילוסופיה"‬
‫מאת‬
‫יואב אבניאון‬
‫הוגש לסינאט אוניברסיטת בן גוריון בנגב‬
‫‪ 81‬מרץ ‪3182‬‬
‫ט"ז אדר ב' תשע"ד‬
‫באר שבע‬