Faculty of Science
TRACKING MOVEMENT IN OVERWINTERING SONGBIRDS: AN RFID APPROACH
2016 | JACKSON WILLIAM KUSACK

B.Sc. Honours thesis

TRACKING MOVEMENT IN OVERWINTERING SONGBIRDS: AN RFID APPROACH
by
JACKSON WILLIAM KUSACK

A THESIS SUBMITTED IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
BACHELOR OF SCIENCE (HONS.)
in the
DEPARTMENT OF BIOLOGICAL SCIENCES
(Animal Biology)

This thesis has been accepted as conforming to the required standards by:

Matthew Reudink (Ph.D.), Thesis Supervisor, Dept. Biological Sciences

Nancy Flood (Ph.D.), Co-supervisor, Dept. Biological Sciences

Mark Paetkau (Ph.D.), Examining Committee member, Dept. Physical Sciences

Mark Rakobowchuk (Ph.D.), Examining Committee member, Dept. Biological Sciences

Dated this 6th day of July, 2016, in Kamloops, British Columbia, Canada

© Jackson William Kusack, 2016

ABSTRACT
Globally, habitat loss and changes in land use are causing an increase in habitat
fragmentation, which can negatively impact some songbird species. Urbanization in particular can
create highly fragmented habitats with landscape corridors that vary greatly in their propensity to
be crossed. In the study of animal movement, there are several factors that limit the quantity and
quality of data that can be collected on individuals. In birds, movements occur quickly and often
high in trees or in heavy brush, making detailed observations of the movements of known, colourbanded, individuals challenging. I utilized a Wi-Fi enabled radio-frequency identification (RFID)
bird-feeder system in order to passively and autonomously track movement events of banded
permanent resident songbirds (i.e., House Finches, Song Sparrows, Dark-eyed Juncos, and
Mountain Chickadees) in the area of Kamloops, BC, Canada. I tracked banded and RFID tagged
songbirds over a 63 day period from 1 January 2016 to 3 March 2016, recording 21732 visitation
events by 28 individuals. From these visitation events I determine 817 movements by House
Finches (n = 23 individuals), 176 movements by Mountain Chickadees (n = 2 individuals), and 15
movements by individuals of other songbird species. I then used ArcGIS to create resistance
landscapes to ask whether movement patterns were best predicted by the “resistance” characteristic
of the habitat between feeder locations or by straight-line distance. The movement patterns
exhibited by tagged birds were not predicted by any of straight-line distance, least-cost distance,
straight-line pathway resistance, or least-cost pathway resistance. However, feeders with a higher
proportion of shrubs and trees were visited marginally more frequently. I also found that females
traveled between feeders more frequently than males.

Thesis Supervisor: Associate Professor Matthew Reudink
ii

ACKNOWLEDGEMENTS
This research would not be possible without the help and supervision of Dr. Matthew
Reudink, and Dr. Nancy Flood. Thanks to Jacob Bailey, Erica Bonderud, Dana Eye, Stephen Joly,
Kristen Marini, Kile McKenna, Cara Snell, and the rest of the BEaC lab for countless hours of
field work. A very special thanks to Dr. Steffi Lazerte for providing R code, Dr. Mark Paetkau for
helping with all electronic issues, and Dr. David Hill for helping with ArcGIS analyses. Funding
for this project was provided by Dr. Reudink’s NSERC grant.

iii

TABLE OF CONTENTS
ABSTRACT ................................................................................................................................... ii
ACKNOWLEDGEMENTS ........................................................................................................ iii
INTRODUCTION......................................................................................................................... 1
MATERIALS AND METHODS ................................................................................................. 6
Radio-Frequency Identification System ................................................................................. 6
Feeder Placement ...................................................................................................................... 8
Study Species ............................................................................................................................. 9
Resistance Classification ........................................................................................................ 10
Data Management ................................................................................................................... 11
Resistance Pathing .................................................................................................................. 11
Feeder Microhabitat ............................................................................................................... 13
Analysis .................................................................................................................................... 13
RESULTS .................................................................................................................................... 14
Visitation Summary ................................................................................................................ 14
Movement Summary .............................................................................................................. 15
House Finch Movements ........................................................................................................ 16
DISCUSSION .............................................................................................................................. 19
Movement Comparisons ......................................................................................................... 19
Dominance and Gender .......................................................................................................... 20
Interspecific Comparisons...................................................................................................... 21
Microhabitat Composition ..................................................................................................... 22
Future Directions .................................................................................................................... 23
Conclusions .............................................................................................................................. 24
LITERATURE CITED .............................................................................................................. 25

iv

LIST OF FIGURES
Figure 1. a) General schematic of feeder design, b) photograph of House Finch eating from an
RFID feeder (photos by J. Kusack). pg 6
Figure 2. Photographs of internal components of an RFID feeder: a) RFID logger b) Raspberry
Pi 2.0 unit (photos by J. Kusack). pg 7
Figure 3. – TRU campus aerial photograph showing feeder locations (aerial photography
obtained from http://www.kamloops.ca/maps). pg 9
Figure 4. – Example least-cost pathway (orange) across a rasterized landscape, in ArcGIS,
between two feeder pairs. Colours represent landscape feature class. Blue corresponds to
buildings, dark green is trees/shrubs, light green is general open space, light grey is roadways
and parking lots, and dark grey is pathways. pg 12
Figure 5. – Summary map displaying movements of a) House Finches (n = 23), b) Mountain
Chickadees (n = 2), c) Dark-eyed Junco (n = 1), and d) Song Sparrow (n = 1). Lines between
fixed feeder locations (i.e., black dots) vary in colour and width. As indicated by the lower
right-hand scale, darker red lines indicate more movement events; Width increases as
redness deepens. Circles around feeder locations also vary in colour and radius. As
indicated by the right-hand scale, darker red circles indicate a higher average feeding time
per individual; Radius increases as redness deepens. pg 15
Figure 6. – Barchart displaying the differences in mean pathway usage (i.e., number of
movements by individual birds) among the different pathways. Error bars show standard error
from mean. pg 16
Figure 7. – There was no relationship between the path usage and straight-line distance (a) or
least-cost distance (b). Points represent individual birds on specific pathways. pg 17
Figure 8. – There was a trend (p = 0.057) towards higher visitation at feeders with a higher
proportion of the microhabitat surrounding the feeder covered by trees and shrubs. Points are
labelled with feeder ID. pg 18
Figure 9. – Female House Finches move among feeders more often (p = 0.02) than male House Finches.
Bars show mean total movement events for each individual. Error bars show standard error from mean. pg
19

LIST OF TABLES
Table 1. – Summary table of visitation events showing: total visits (i.e., the total number of
visitation events among all individuals), mean (i.e., the mean number of visits per individual),
and n (i.e., number of individuals recorded). pg 14

v

INTRODUCTION
Globally, habitat loss and changes in land use are two of the greatest threats to biodiversity
(Robinson et al. 1995). Based on information from satellite imaging, North America experienced
the highest rate of gross deforestation in the last decade compared to any other continent (Hansen
et al. 2010). At the national level, Canada experienced the second largest rate of gross
deforestation, exceeded only by Brazil (Hansen et al. 2010). In addition to direct habitat loss,
deforestation can also lead to habitat fragmentation, which occurs when natural habitat is broken
into patches and the distance between patches is increased (Andrén 1994). Fragmentation is
strongly associated with habitat loss, and while it is difficult to isolate the effects of fragmentation
from those of loss, fragmentation by itself can increase edge effects and negatively impact natural
movement patterns of animals (Fahrig 2003).
Habitat fragmentation leads to a change in spatial landscape structure, which can negatively
impact songbird species, especially when fragmentation causes overall habitat area to decrease
(Betts et al. 2006). The size of the gaps in continuous habitat, whether anthropogenic (e.g., roads
and railways) or natural (e.g., rivers), negatively affects songbird movements between habitat
patches (Collinge 1996; Tremblay and St Clair 2009). A study conducted on forest songbirds in
Calgary Alberta, Canada showed that the size of the gaps between forest patches was the most
important factor deterring songbird movement (Tremblay and St Clair 2009). Some songbird
species will move longer distances (up to 2-3x longer than the straight-line distance) through
connected habitat in order to avoid crossing gaps (Desrochers and Hannon 1997). Similar trends
have been found in gliding mammal species. Smith et al. (2013) found that in some cases flying
squirrels (Glaucomys sabrinus) would detour around the gap even in instances where the gap was
6.8 times shorter than the detour path (Smith et al. 2013). If fragmentation limits movement, it has
the potential to limit gene flow (i.e., movement of alleles between populations), reduce access to
resources, reduce access to habitat, increase competition, and limit ability to avoid predators
(Fahrig 1998). The effect of gaps on landscape level movement can be examined by studying
habitat connectivity (i.e., facilitation of movement between isolated patches) (Taylor et al. 1993).
In the study of animal movement, there are several factors that limit the quantity and quality
of data that can be collected on individual movements. In birds, movements occur quickly and
1

often high in trees or in heavy brush, making detailed observations of the movements of known,
colour-banded, individuals challenging. Technologies exist that provide high resolution for
tracking individual movements, but each has its own drawbacks. Satellite tracking systems
utilizing GPS coordinates can follow the movements of an individual with high resolution, but are
often prohibitively expensive and the transmitting devices are relatively large (>22g) so that they
are not suitable for use on small animals, such as songbirds (Bridge et al. 2011). While smaller
satellite relay devices exist, they are limited to providing only a few data points with decreased
resolution of location, which limits remotely transmitting GPS studies to larger birds (Bridge et al.
2011). In contrast, light-level geolocators are small, lightweight devices that can be placed on
songbirds to track migratory movements across the globe, but their resolution is limited to ~200km
(Bridge et al. 2011). Similarly, GPS data loggers are now small enough to attach to songbirds, but
due to battery requirements, are only able to record a handful of GPS locations at pre-programmed
times. Both light-level geolocators and GPS data loggers also require subsequent recapture of
individuals to retrieve and download the movement data. Finally, radio-tracking provides
extremely high-resolution spatial data on the movements of even small songbirds (Sokolov 2011).
However, radio-tracking is limited by the number of individuals that can be tracked at a given
time, the life of the battery in the transmitting device (several weeks) as well as the time and effort
involved in tracking individuals with an antenna. Thus, although all of these techniques allow us
to measure movement events, each also has limitations in terms of data resolution, collection effort,
cost, and feasibility.
Radio-frequency identification (RFID) technology can be used to monitor small-scale
movements by small birds across a landscape (Bonter and Bridge 2011). Radio-frequency
identification is a method of wireless identification, recently adapted by the ornithological
community, that utilizes electromagnetism to transfer data from passive integrated transponder
(PIT) tags into a reader (Bonter and Bridge 2011; Bridge and Bonter 2011). Using this approach,
objects or animals with PIT tags attached can be identified if they come within close proximity of
the electromagnetic field generated by the RFID reader’s antenna. This system is widely used in
industry to track the movement of assets, products and employees (Violino 2005). The technology
is utilized to ask three simple questions: who/what is being identified, where is it being identified,
and when is it being identified? While the data transfer and information collected is relatively
simple, the benefit of RFID technology is that it allows for real-time data transfer with no human
2

intervention (Violino 2005). The power of RFID systems is in the accumulation of temporally
sequential tagging events, allowing for a log of movements to be created.
There are several advantages to using RFID to track animal movements. First, data can be
collected autonomously once a bird is banded with a PIT tag. This is especially pertinent to
ornithological studies in which repeated capture can cause stress to the individuals being studied.
Second, PIT tags vary in size from 8mm to 34mm and can weigh less than 0.1g, allowing for
various methods of attachment and implementation with even very small study species, such as
hummingbirds (Brewer et al. 2011; Hou et al. 2015) and bees (Schneider et al. 2012). Third, PIT
tags have no internal power source, which effectively allows them to remain viable indefinitely.
The power source for PIT tags is provided by the electromagnetic field of the RFID antenna,
allowing them to transmit data with no internal power source. The final benefit is the affordable
pricing of the tags and readers. PIT tags cost under $5 USD per tag, and the reader boards cost
under $50 USD for a prebuilt unit (Bridge and Bonter 2011). Compared to the hundreds or
thousands of dollars cost for each satellite (i.e., relay device) or GPS logger, RFID is an
inexpensive option for tracking movements.
There are, however, several limitations to the use of RFID technology to monitor avian
movements. Individuals must land within range of an antenna to be recorded, and thus movement
between antenna locations can only be registered as a straight-line movement, although in reality
an individual may take a circuitous path. In addition, this technique is only effective when animals
go to, or can be enticed (e.g., by providing food) to, known locations where the antenna is located.
Nevertheless, the power of RFID systems is in the ability to autonomously collect thousands of
data points on the locations of individuals over extended periods of time; from this one can then
infer movement patterns and examine how landscape features, such as habitat fragmentation,
influence individual movements.
The study of the effects of landscape patterns and habitat fragmentation on movement can
be approached in several different ways. One method of examining movement within a landscape
is to look at least-cost pathing. This method of estimating distance between resources patches is
readily available as a standard ArcGIS (Esri 2005) application and has been used to examine
ecological questions involving habitat connectivity, and movement patterns (e.g., Graham 2001;
McRae and Beier 2007). From a biological standpoint, least-cost pathing examines a landscape
3

from the perspective of a single individual moving between two patches in a landscape. The leastcost path is the route that poses the least resistance, or best facilitates movement (Adriaensen et al.
2003). In the case of a bird crossing a fragmented landscape, the least-cost path would be the one
that makes flight easiest. Examples of factors that could influence movement resistance/facilitation
include exposure to predators, or presence of physical barriers.
Circuit theory has also been applied to the study of movement. Circuit theory examines
movement and connectivity across a landscape by examining connections between cells in a
rasterized (i.e., composed of cells) landscape (McRae et al. 2008). In the rasterized landscape, each
cell represents a node and each pair of nodes is connected by a resistor. Similar to electricity
passing through a circuit, these nodes and resistors represent ecological variables relevant to
facilitating or hindering movement. Circuit theory examines the landscape as a whole and models
movement between two points to determine important corridors and pinch-points (i.e., small areas
with lots of movement) for movement (McRae et al. 2008). Both circuit theory and least-cost
pathing provide models that allow us to predict movement patterns, and with RFID data collection
I can examine model appropriateness. Although circuit theory provides a unique modelling
approach to habitat connectivity and movement, I was unable to attempt circuit theory modelling
in this analysis.
RFID data can be used in conjunction with movement/connectivity models in several ways,
including: 1) creating a resistance landscape, or 2) examining the appropriateness of predictive
movement models. Resistance landscapes are geographic representations of the resistance to
movement of the different components of a particular landscape and are commonly used in
conjunction with least-cost pathing or circuit theory to examine movement. To create resistance
maps, resistance values must be assigned to each landscape feature. Most often resistance values
are based on expert opinion and are determined via observations (Zeller et al. 2012). RFID data
consist of a series of sequential point counts and thus RFID is a prime candidate to be used as
relocation data (i.e., data on individuals recorded in different areas at different times), which are
necessary for creating a resistance map and assigning resistance values (Zeller et al. 2012). Using
geographical modeling, the frequency of certain movement events, and categorized landscape
composition, resistance values for each landscape feature can be determined (Zeller et al. 2012).
Due to the limitations of this study, this approach was not used to determine my resistance
landscape and the most appropriate resistance values, although I recognize it’s potential. RFID
4

data can also be used to examine the appropriateness of predictive movement/connectivity models.
Although RFID data does not contain information pertaining to actual movement pathways,
sequential RFID logs from the same individual show a movement has occurred. Movement
frequencies can be related to different movement/connectivity models to examine which better
predicts actual movement events. In other words, which theoretical model for preceding movement
events is closest to reality?
Here I use RFID-based tracking to examine movement patterns in songbird species
traversing an urbanized and highly fragmented environment. I utilized a Wi-Fi enabled RFID birdfeeder system implemented within the Thompson Rivers University (TRU) campus, Kamloops
B.C., in order to passively and autonomously track movements of certain banded permanent
resident songbirds (e.g., House Finches [Haemorhous mexicanus], Song Sparrows [Melospiza
melodia], Dark-eyed Juncos [Junco hyemalis], and Mountain Chickadees [Poecile gamelii]).
Focusing on House Finches, I asked whether movement patterns across the landscape were
predicted by two models: straight-line pathing and least-cost pathing. These models represent two
potential movement patterns. Least-cost pathing relies on movement being driven by resistance
values in the landscape, while straight-line pathing ignores resistance, thus determining pathway
based on a direct line between points (Adriaensen et al. 2003). Therefore, whichever model best
predicts movement determines whether resistance plays a role in movement events of songbirds in
a fragmented landscape. Within these two models I examined the effects of straight-line distance
(i.e., the direct distance between two points), least-cost distance (i.e., the distance along the path
of least resistance between two points), straight-line resistance (i.e., a sum of the resistance values
along a direct path between two points), and least-cost resistance (i.e., a sum of the resistance
values along a path of least resistance between two points) on movement. I predicted that House
Finches would be most sensitive to crossing buildings and roadways, and movement patterns
would be best predicted by the least-cost modeling, either distance or resistance, thus showing that
House Finch movement patterns are affected by landscape resistance. Additionally, since House
Finches of different age and sex differ in behavioral dominance hierarchical position (Brown and
Brown 1988; McGraw and Hill 2000), which may affect access to feeders thus creating the need
to move across a landscape to access resources, I asked whether movement patterns differed among
age and sex classes. Finally, to predict feeder visitation and whether it is driven by landscape
features directly around the feeder (i.e., microhabitat) I also asked whether visitation frequency
5

was associated with surrounding microhabitat and whether movements were predicted by feeder
usage (i.e., were paths between high-use feeder pairs traversed more frequently).
MATERIALS AND METHODS
Radio-Frequency Identification System
My radio-frequency identification (RFID) system was composed of four key components:
a passive integrated transponder (PIT) tag, a RFID antenna, a RFID logger, and a Raspberry Pi 2.0
unit (a small computer programmed with Python code). I obtained 10mm and 12mm in length
hermetically sealed glass passive integrated transponder tags from Cyntag RFID systems
(Kentucky, USA). I obtained RFID reader boards from Cellular Tracking Technologies
(Pennsylvania, USA).
The RFID components were contained within a bird feeder made of an approximately 2.5
foot piece of PVC pipe, which acted as both a reservoir for black oil sunflower seed and a physical
barrier protecting the internal circuitry. Within the pipe, several 3D-printed plastic pieces
supported the RFID system. A single hole was drilled in the front of the feeder allowing birds
access to the seed while sitting on an outside perch. In order to give the feeder the appearance of
a food source, a small window was cut in the outer PVC shell, and covered with clear plexiglass.
This kept the internal cavity watertight while allowing the perching birds to see the food inside
(refer to Figure 1).
a)

b)

Figure 1. a) General schematic of feeder design, b) photograph of House Finch eating from an
RFID feeder (photos by J. Kusack).
6

PIT tags are temporarily powered when in an electromagnetic field generated by the RFID
antenna. The RFID antenna cycles its electromagnetic polling every 1600 milliseconds when
scanning for PIT tags. This polling frequency optimizes the potential for recording every visitation,
while avoiding continuous polling, which would create unnecessary data loads. When temporarily
powered, a PIT tag broadcasts its unique 10 digit alphanumeric identity and the RFID antenna
relays this information to the RFID logger (Figure 2a). The logger stores a time stamped record of
an individual’s visitation on a SD card and relays the information to the Raspberry Pi 2.0 unit. The
Raspberry Pi 2.0 unit (Figure 2b) then relays the information wirelessly over a Wi-Fi network so
that it can be stored in a server database.

a)

b)

Figure 2. Photographs of internal components of an RFID feeder: a) RFID logger b) Raspberry
Pi 2.0 unit (photos by J. Kusack).
We attached PIT tags to individuals of my study species using custom-made leg bands with
sleeves to hold the PIT tags. These bands were made of 3.3mm catheter tubing, which made them
7

flexible enough to prevent snapping during the banding process. Because the legs of study species
varied in thickness (i.e., required different sized leg bands) the tubing was large enough in diameter
to fit all leg sizes, but small enough to prevented bands from falling off birds with the smallest leg
size. To meet the requirements of my animal care and banding protocols, these leg bands had to
weigh less than 5% of the mass of a banded bird. Tags weighed 0.25g (10mm tags) and 0.30g
(12mm tags), which was less than 3% of an individual’s body weight for every bird in the study,
with the exception of Mountain Chickadees. Although the 12mm tags weighed less than 5% of a
chickadee’s mass, 10mm tags were used for this species because they weighed less than 3% of an
individual bird’s weight.
This system enables the recording of data points, each of which consisted of three
components: a who, a when, and a where. The “who” component of a data point is provided by
the unique PIT tag attached to each individual in the study. Each visit to a recording feeder is
paired with a unique tag that provides information about who was recorded. The “when”
component of the data point is provided by the time stamp created when an individual visits a
recording feeder. The “where” component is provided by the fixed location of the RFID feeders.
Because each visit is made to a single feeder, feeder ID can be compared with feeder location to
determine where the visitation event occurred.
Feeder Placement
RFID enabled feeders were placed in six fixed locations on and near the TRU campus.
These locations were selected based on their potential for access to a power source and Wi-Fi
connection, suitability for feeder hanging, and seclusion from human activity. RFID feeders were
placed in six locations: one at the west end of the Ken Lepin Building (labelled 2400), two
immediately south of the Ken Lepin Building (labelled 2700 and 2100), one at the Children’s
Therapy & Family Resource Center across McGill road from the TRU campus (labelled 2200),
one east of the Culinary Arts Building (labelled 2300) and one behind Cplul’kw’ten, the First
Nations Gathering Place (labelled 1500).

8

Figure 3. – TRU campus aerial photograph showing feeder locations (aerial photography
obtained from http://www.kamloops.ca/maps).
Study Species
Target species for this study were feeder-habituated birds species found on the TRU
campus: House Finches, Song Sparrows, Dark-eyed Juncos, and Mountain Chickadees. These
species are considered permanent residents in the Kamloops area and are found on campus
continuously throughout winter.
We caught birds using 38mm mesh, 4-shelved, polyester mist nets placed between perching
locations in close proximity to the feeders. Mist nets were constantly monitored over the period
that they were open by one or more observers. When a bird was caught in the net, it was removed
as quickly as possible and placed into a cotton bag with a draw string closure to prevent escape.
The bag provided shelter from cold temperatures and wind, and additionally acted as a visual
barrier between the bird and the processor, thus reducing any stress that might be caused by
proximity to humans.

9

Study species were banded with a single aluminum Canadian Wildlife Service-issued band,
as well as the custom-made PIT tag band. Birds caught before 27 October 2015 had their PIT tag
bands secured using black electrical tape, or coloured electrical tape in the case of Mountain
Chickadees, which aided visual identification of individual chickadees. After this date, I applied
viscous Gorilla Super Glue to the electrical tape before application in order to reduce the potential
of PIT tags falling off in the field. Care was taken to apply the smallest amount of super glue
possible to minimize the potential for glue coming in contact with the banded bird’s leg. Three
birds were re-caught birds in the study that were originally banded before the gluing date, and all
three had lost their PIT tags, leading me to believe that the original methodology was insufficient
to attach the tag. No other birds were re-caught and therefore no assumption can be made about
the subsequent procedure.
During banding, the following data was recorded for each species: date of capture, location
of capture, PIT tag ID, aluminum band ID, species, sex, wing length, tarsus length, tail length, and
weight. A frontal photograph of the individual was taken post banding.
Resistance Classification
In order to examine movement patterns across an urban landscape, a resistance map was
created in ArcGIS (Esri 2015). First, I created shape files using polygons (i.e., enclosed shapes) to
represent the landscape features found on the TRU campus: buildings, trees/shrubs, roadways,
pathways, and general open space. I then combined these polygon files into a single shape file and
converted it into a raster file (i.e., a grid of cells) representing the entire landscape. Next, I used
the weighted overlay tool, which assigns a factor of resistance to each landscape feature in the
shape file. I assigned the following resistance factors: trees/shrubs – 1, general open space – 3,
pathways – 6, roadways – 7, buildings – 9. These values were chosen based on the degree to which
each of the landscape feature was assumed to limit movement by songbird species. Because this
tool limits resistance factors to be an integer between 1 and 9, the specificity of the resistance
values was limited. In order to maximally represent the range of resistance values, trees/shrubs
were assigned the lowest resistance value of 1 and buildings were assigned the highest resistance
value of 9. Subsequently the remaining resistance values were assigned based on their assumed
intermediate effect on movement. Trees and shrubs provide areas of shelter from vision, thus
providing a pocket of potential sanctuary from predators and other dangers. These sanctuaries
10

should facilitate movement and therefore receive the lowest resistance value. On the other end of
the spectrum, buildings exist as large exposed structures that act as focal points for human activity.
For this landscape feature, there are no areas of refuge, the building may act as a physical barrier
and birds may avoid the area due to high abundance of human occurrence.
Data Management
The RFID system outputs uniquely identified, time-stamped data points that are then
modified into visitation events. This modification step is necessary because of the polling
parameters of the system. The polling cycle for the system is 1600 milliseconds and will continue
to record an individual PIT tag within the field for the duration of its visit. This results in a series
of concurrently recorded data points, 1600 milliseconds apart, for the same individual. To clean
up the raw output, any series of concurrent data points at a single feeder by a unique individual is
simplified into a single visitation event using a script created in the statistical program R (R Core
Team 2016).
In addition to producing the data describing visitation events, R scripts were used to
determine movement events. A movement event was calculated from sequential visitation events
by a given individual. For example, if a bird (e.g., 0620000514) was recorded visiting one feeder
(e.g., 2100) and then the same bird was recorded at a separate feeder (e.g., 2200) an hour later, a
movement occurred. Using this approach, movement frequency between any pair of feeders could
be determined.
Resistance Pathing
For straight-line distance, I measured the distance between feeder pairs using the measure
tool in ArcGIS. To determine the least-cost pathway between a pair of feeders, the feeder at which
the movement originated (i.e., the source feeder) was input into the cost distance tool. This
produced a cost distance raster and a backlink raster. The cost distance and blacklink rasters are
geographic information files portrayed in raster format (i.e., objects shown as a rectangular grid
pattern rather than normal shapes). These files contained information pertaining to the cost of
movement to any cell away from the source feeder, taking into account distance moved from the
source feeder and the resistance values of the landscape. These raster files were then input, along
with a destination feeder shape file, into the cost path tool, which outputs a least-cost pathway
11

raster. This raster represents the pathway of least resistance from the source feeder to a single
destination feeder. This method was repeated for every feeder to create least-cost pathways
between all possible pairs of feeders (n = 15). I then converted the raster pathways, which are
portrayed as a rectangular grid pattern, to polylines (Figure 4), which are shown as solid lines, and
calculated their distances using the calculate geometry function.

Figure 4. – Example least-cost pathway (orange) across a rasterized landscape, in ArcGIS,
between two feeder pairs. Colours represent landscape feature class. Blue corresponds to
buildings, dark green is trees/shrubs, light green is general open space, light grey is roadways
and parking lots, and dark grey is pathways.
For each of the calculated pathways representing a least-cost route between a feeder pair,
I used the cost distance raster of the source feeder to determine the cumulative resistance along
that path. To do this, I used a tool in ArcGIS called extract values to points. Because the cost
distance raster contains information about the cost of movement to any pixel away from the source
feeder, the resistance value for least-cost movement to any destination feeder can be determined.
Thus, using the cost distance tool I was able to extract a resistance value for each feeder pair. To
calculate the cumulative resistance value for the straight-line pathway, I used ArcGIS to sum up
the resistance values for each land use pixel crossed.

12

Feeder Microhabitat
To classify the microhabitat around each feeder, I grouped the general types of habitats
within the study area into five categories: trees, shrubs, buildings, pavement, and open space. I
used a 10m buffer around each feeder to represent the microhabitat affecting the feeder. I then
mapped the buffer zone on satellite images, approximating the proportion of the area covered,
when viewed aerially, by trees, shrubs, buildings, pavement and grass/dirt. These cover areas were
then imported into ArcGIS by creating shape files. To determine the proportion of cover for each
habitat type within the 10m buffer, I divided the area of each of the created polygons (i.e., trees,
shrubs, buildings, etc.) by the total area of the buffer zone. Using this method, I approximated the
habitat composition in a 10m buffer around each of the RFID feeders. This was used to compare
differences in the microhabitats among the feeders and to determine if visitation was driven by
habitat composition.
Analysis
For all analyses, I used a 63 day period starting on 1 January 2016 and ending on the March
3 2016, and due to sample size restrictions, limited the analyses to House Finches, unless otherwise
stated. All statistical analyses were performed in JMP 12 statistical software (JMP 2016). To
examine differences in movements among House Finches among the different feeder pairings (i.e.,
feeder pathways), I ran a generalized linear mixed model accounting for individual ID as a random
effect.
To examine if any distance or resistance factors predicted pathway use, I constructed
generalized linear mixed models with pathway use (i.e., the number of times a path is used
regardless of direction) as the response variable, individual birds nested within path ID as a random
effect, and separate models with straight-line distance, least-cost distance, straight-line resistance,
or least-cost pathway resistance as a main effect. By nesting birds within path ID, it pairs all
possible pathways with any individual bird IDs that moved along that pathway. Accounting for all
pairings of bird ID/path ID as a random effect reduces the possibility that statistical significance
is due to any random effects of bird and pathway interactions.
I ran a series of linear regressions to examine whether visitation frequency, microhabitat
composition, or visitation proportion (i.e., the relative proportion of visits that occurred at that
13

feeder) predicted path use. I compared the relationship between microhabitat composition and
visitation frequency by running a series of linear regression analyses using proportion of
trees/shrubs and proportion of buildings as predictor variables and visitation frequency as a
response variable. To see if movement between feeder pairs was driven simply by the number of
visitations at the two feeders combined, I examined the relationship between path use and the
proportion of the total visitation events that occurred at the two feeders. Proportion of total
visitation events was calculated by adding the total visitation events at both feeders, and dividing
this value by the total visitation events at all feeders combined.
To examine whether pathway use differed by age or sex, I constructed a mixed model with
individual ID as a random effect, age and sex as main effects and pathway use as a response
variable.
RESULTS
Visitation Summary
Over a period of 63 days, I recorded 21732 visitation events by 28 individuals. These
visitation events were spread among 23 House Finch individuals, 2 Mountain Chickadee
individuals, 2 Dark-eyed Junco individuals, and 1 Song Sparrow (Summarized in Table 1).
Species
House Finch
Mountain Chickadee
Dark-eyed Junco
Song Sparrow

Total visits
18012
1904
759
1057

Mean
783.1
952.0
379.5
1057.0

n
23
2
2
1

Table 1. – Summary table of visitation events showing: total visits (i.e., the total number of
visitation events among all individuals), mean (i.e., the mean number of visits per individual),
and n (i.e., number of individuals recorded).
In House Finches, mean visitation frequency was higher in females than in males (mean
visits per female = 1540 ± 260 SD, n = 8, mean visits per male = 283 ± 220, n = 12, t = -3.69, P =
0.002). On average, female House Finches visited feeders more frequently than males.

14

Movement Summary
Over the same period, I recorded 817 movements made between feeders by 23 banded
House Finches (Figure 5a), 176 movements made by two Mountain Chickadees (Figure 5b), 11
movements made by a single Dark-eyed Junco (Figure 5c), and four 4 movements by a single Song
Sparrow (Figure 5d).
a)

b)

c)

d)

Figure 5. – Summary map displaying movements of a) House Finches (n = 23), b) Mountain
Chickadees (n = 2), c) Dark-eyed Junco (n = 1), and d) Song Sparrow (n = 1). Lines between
fixed feeder locations (i.e., black dots) vary in colour and width. As indicated by the lower
right-hand scale, darker red lines indicate more movement events; Width increases as
redness deepens. Circles around feeder locations also vary in colour and radius. As
indicated by the right-hand scale, darker red circles indicate a higher average feeding time
per individual; Radius increases as redness deepens.

15

House Finch Movements
There were significant differences in mean pathway usage (i.e., average number of
movements among all individuals) among the 15 possible pathways (Figure 6; Mixed Model; n =
345, F14,308 = 3.30, p < 0.0001). For example, the pathway between feeder 2100 and feeder 2300
(i.e., 2100_2300) was travelled more frequently than the pathway between feeder 2200 and feeder
2400 (i.e., 2200_2400).

Figure 6. – Bar chart displaying the differences in mean pathway usage (i.e., number of
movements by individual birds) among the different pathways. Error bars show standard error
from mean.
There was no relationship between the straight-line distance between pairs of feeders and
the frequency with which individuals moved among these feeder pairs (Figure 7a; Mixed model;
n = 345, F1,343 = 0.26, p = 0.61). That is, shorter pathways weren’t used more—or less—often
16

than longer ones. There was also no relationship between least-cost distance between feeder pairs
and movement frequency (Figure 7b; Mixed model; n = 345, F1,343 = 0.29, p = 0.59); what I had
labelled the “easiest” routes to travel weren’t used more or less often than the more difficult ones.
Neither least-cost pathway resistance (Mixed Model; n = 345, F1,343 = 1.61, p = 0.21) nor straightline pathway resistance (Mixed Model; N= 345, F1,343 = 0.96, p = 0.33) predicted path use.
Pathways with lower least-cost or straight-line resistance were not used more often than pathways
with higher resistance values.
a.

b.

Figure 7. – There was no relationship between the path usage and straight-line distance (a) or
least-cost distance (b). Points represent individual birds on specific pathways.
There was a trend towards higher visitation rates at feeders with a higher proportion of the
area in the immediate vicinity of the feeder covered by trees/shrubs (r2 = 0.64, p = 0.057; Figure
8). There was no relationship, however, between visitation rate and the proportion of the
microhabitat surrounding the feeder covered by buildings (r2 = 0.36, p = 0.20). Feeders with
buildings in close proximity were not visited more—or less—often that feeders further from
buildings. In addition, there was no relationship between path use and visitation proportion at the
feeder pair the birds moved between (r2 = 0.06, p = 0.36). In other words, pathways between feeder
pairs that had high visitation rates were not traversed more frequently than pathways between less
visited feeders.

17

1500

2300

2400

2100

2700
2200

Figure 8. – There was a trend (p = 0.057) towards higher visitation at feeders with a higher
proportion of the microhabitat surrounding the feeder covered by trees and shrubs. Points are
labelled with feeder ID.
There was no relationship between movement path use and combined feeder pair visitation
proportions (r2 = 0.064, p = 0.36). When I examined whether movement patterns differed among
age and sexes, there was a marginal, but non-significant trend towards more movements being
made by adult (after hatch year/after second year) birds than younger ones, and a significant effect
of sex, with females moving among feeders more than males (Figure 9; age: F1,17 = 3.67, p = 0.08;
sex: F1,17 = 7.07, p = 0.02).

18

Figure 9. – Female House Finches move among feeders more often (p = 0.02) than male House Finches.
Boxplots show quartiles, median, and hinges show upper and lower limits.

DISCUSSION
Movement Comparisons
We observed differences in movement frequency among pathways, suggesting that some
biotic or abiotic factors were influencing individual movements. Surprisingly, however, no
measure of distance or resistance predicted path usage. Previous studies have found that gaps in
habitat, whether anthroprogenic or natural, can inhibit movement (Desrochers and Hannon 1997;
Tremblay and St Clair 2009; Smith et al. 2013). House Finches exhibited no detectable resistance
to movement from landscape features, despite the presence of certain features (e.g., roadways,
parking lots) that acted as gaps in vegetative habitat within my study landscape. One predictor of
gap crossing behaviour is detour efficiency (Smith et al. 2013). As the energy required to move
around a gap decreases, and/or the risk of predation while moving through a gap increases, there
reaches a point where it is safer and more efficient to go around than to cross the gap. Because of
the limitless access to black oil sunflower seeds provided to the birds in this study, energy

19

expenditure in flight might not have been a high priority when traversing the landscape, especially
considering the short distances between feeders (<600m). Thus, energy efficiency issues may be
unlikely to explain the pattern of House Finch movements observed here. If this is true, straightline distance should have predicted movement frequency; however, there was no relationship
between the distance between feeders and path usage. Thus, the lack of a relationship between
movement and distance/resistance may indicate that other factors, such as the presence of vehicles,
humans, or predators, none of which were measured in this study and thus weren’t considered in
my models are more important than distance and resistance measures. Previous studies have shown
that increased movement and dispersal increases risk of predation (Yoder et al. 2004). If a predator
was within the study area during any period of the recorded movement events, individual
movement events may be reduced to lower the risk of predation. Additionally, vehicular traffic
can reduce movement frequency (Tremblay and St Clair 2009).
Another potential issue confounding the results of least-cost analysis are my model
resistance value parameters. These values may not realistically represent the resistance values of
the landscape features found on a university campus, leading to incorrect model parameters. When
using least-cost path analysis, choosing appropriate resistance values for landscape features is the
most limiting step; although the values should accurately describe the resistance to movement,
they may not do so (Adriaensen et al. 2003). Most often, resistance values are based on expert
opinion, rather than empirical evidence or modelling (Zeller et al. 2012). Additionally, when using
readily available ArcGIS tools to apply resistance values, the values are limited to integer values
from 1-9, and may be inaccurate for this reason as well. For future studies, RFID movement data
could be analyzed to produce a resistance map (e.g., using the methods in Zeller et al. 2012), which
would enable us to infer which landscape features most influence movement patterns.
Dominance and Gender
Dominance relationships may also influence movement patterns. Previous research on
House Finches

has

demonstrated that

plumage colouration is related

to male condition

and quality. More specifically, colouration is related to nutritional condition and parasite
load during moult (Hill

and

Montgomerie

1994;

Thompson

et

al.

1997),

indicating that carotenoid colouration is an honest indicator of individual health. Despite being an
indicator of individual health, outside of the breeding season plumage colouration in male house
20

finches seems to be negatively correlated with indicators of behavioral dominance (Brown and
Brown 1988; McGraw and Hill 2000); less colourful male House Finches are able to gain greater
access to feeder resources during the winter months (Brown and Brown 1988). Although many
studies have looked at behavioral dominance in relation to plumage colouration, it is still unclear
how colour and dominance relate to movement patterns and landscape use. Colouration could be
influencing movement patterns through dominance relationships. Because females and less
colorful males are dominant over colourful males during the winter months (Brown and Brown
1988; McGraw and Hill 2000) they would be expected to be able to displace brighter males more;
thus leading to colourful males exhibiting the most movement events, as they are forced to move
to obtain resources.
In this study, female House Finches moved between feeders more frequently than males
(independent of male colouration intensity), contrary to expectations. This might be expected if
males were dominant to females, resulting in the need for females to move more often to access
different feeders however, females have higher feeder dominance in the winter. In this case, this
could simply be the result of individual female House Finches exhibiting more visitation events
on average than individual males (mean visits per female = 1540 ± 260 SD, n = 8, mean visits per
male = 283 ± 220, n = 12), which would thus result in more movement events. Because of the
feeder design, only birds that land directly on the perch will be logged by the RFID logger. If
females are exhibiting dominant feeder behaviour, males may be forced to forage for dropped
seeds around the feeder rather than at the feeder itself, resulting in them going unrecorded. Future
feeder design could account for this potential behaviour by adding an additional perch and RFID
antenna for subdominant individuals to land on. This would allow for recording of additional
individuals that may have reduced access, due to dominance interactions, to the single seed access
perch.
Interspecific Comparisons
Closely related species may behave differently when crossing heterogeneous landscape
matrixes (Pither and Taylor 1998; Hokit et al. 1999; Ricketts 2001). For example, the damselflies
Calopteryx maculate and Calopteryx aequabilis experience different resistances when travelling
through forest and pasture areas (Pither and Taylor 1998). C. maculate move more readily through
pasture than through forest, whereas C. aequabilis moves with equal ease through both
21

environments. “Songbirds”, although a broad taxonomic group, have shown different resistance
patterns to movement in variable landscapes. In a study on landscape permeability, translocated
Ovenbirds (Seiurus aurocapilla) and White-throated Sparrows (Zonotrichia albicollis) were found
to exhibit different resistances to movement in different landscape types (Gobeil and Villard 2002).
Agricultural landscapes provided the most resistance to movement for the Ovenbirds, which is
considered a forest specialist, whereas naturally patchy landscape presented the most resistance to
movement in the White-throated Sparrow, which is surprising considering it is a habitat generalist.
House Finches are common urban birds that are habituated to feeders; they may thus be less
affected by the features in an urban landscape matrix than are some other closely related finches,
which may only be present within city limits at certain times of year and exhibit different degrees
of feeder visitation (e.g., Cassin’s Finch [Haemorhous cassinii], Pine Siskin [Spinus pinus],
American Goldfinch [Spinus tristis], Common Redpoll [Acanthis flammea]). In this study, the low
sample size of other tagged species (i.e., two Mountain Chickadees, one Song Sparrow) limited
my ability to compare species with respect to their movement patterns. It would be informative to
have more data from these and other species, but their low density in the study area limits their
detection frequency and probability of being caught during the banding process. Looking at the
recorded movement events from all species (Figures 5-8) I expect that a larger sample size would
result in significant differences in movement patterns among species.
Microhabitat Composition
In areas of greater cover (e.g., shrubs, thickets, and forested areas) urban birds (e.g., House
Finches and House Sparrows [Passer domesticus]) are less likely to give up on foraging at a feeder
with limited resources (Shochat et al. 2004). This is thought to be the result of the tradeoff between
risk of predation and the possible reward food resources. At feeders where there is less risk of
predation (i.e., more shrubs or trees sheltering the feeder), birds are less likely to abandon the
feeder. Here I found a trend (r2 = 0.64, p = 0.057) towards higher visitation proportion at feeders
with a higher proportion of trees and shrubs within 10m of the feeder. This trend supports the
predation threat theory. The non-significance of this trend could be due to the chosen microhabitat
area (i.e., 10m) being too small for an adequate representation of surrounding microhabitat
composition. If the examined microhabitat area were to increase (e.g., 20m or 50m), there is
potential that proportion of trees and shrubs would better predict feeder visitation. For example,
22

feeder 2700 has a low visitation rate but a moderate proportion of trees and shrubs in the 10m
microhabitat zone. If the zone were increased to 50m, it would encompass parts of the Ken Lepin
building, the adjacent parking lot, and a larger portion of the adjacent open grassy area, without
added large portions of trees or shrubs. A 50m microhabitat zone may not accurately represent the
usable area around the feeder during a feeder visit, a larger area may encompass more areas to hide
from predators, which may be important factors in feeder visistation.
Future Directions
There are several modifications that would improve my ability to understand how
landscape patterns influence movement. First, I would deliberately pair feeders, allowing for pairwise analysis of pathway resistance. This would allow for a simple analysis to compare movement
from a source feeder to two choice feeders, one across a path of high resistance and one of low
resistance. Another modification would be to increase the area encompassed by the feeder matrix.
This could be done by adding additional feeders, strategically placing them across the landscape
and spacing them equivalently to cover the maximum area. This would allow us to ask questions
about landscape connectivity on a broader scale. For example, in this study a connection between
feeders may be composed of a few trees, crossing a pathway and crossing gaps of open space. If
this feeder matrix was placed across a city, a connection between feeders may be composed of a
park, several neighborhoods, and crossing several busy streets. While the theory remains the same
for both examples, the questions become more applicable to conservation practices and urban
development planning. One way to achieve this goal is to turn to citizen science and placing RFID
enabled feeders into the backyards of local interested community members. This would allow us
to monitor a larger area, while expanding public outreach for science and conservation.
Next, I would band and monitor more species and more individuals, including closely
related finch species and species that vary in their aversion to human environments. Banding every
individual in an area would provide the clearest picture for any movement study, but because of
the nature of movement (i.e., there is immigration and emigration from the population) this is not
always possible. The addition of Cassin’s Finches, American Goldfinches, Pine Siskins, and
Common Redpolls to the database would allow for comparison of movement patterns among finch
species. Because these species appear in the study area in large numbers only in the winter and my
system works best in winter (due to feeder reliance), banding these species would allow us to
23

increase the amount of movement data collected and thus increase the clarity of movement
analysis. Additionally, because some of these individuals exhibit seasonal short distance migration
patterns, banding individuals in the winter months would allow us to examine seasonal migratory
return rates.
Conclusions
Contrary to predictions, House Finch movements were not predicted by distance or
resistance measures in the least-cost pathing model. This prediction was based on the assumption
that House Finch movements are significantly affected by landscape features such as buildings,
and roadways. Based on these results, House Finches experience no significant effect of landscape
composition (e.g., buildings) on movement patterns. Despite the lack of a relationship between
House Finch movement patterns and landscape resistance, these results show that it is possible to
autonomously and passively record movement patterns in songbirds using an RFID enabled bird
feeder system. This non-invasive system allows for integration into an already established feeder
dynamic with local overwintering songbirds. Establishing a continuously monitoring system that
tracks detections autonomously limits the effects of human observers on the movement patterns,
and allows for a more complete picture of movement across a landscape. Using this powerful
system, data for any type of movement pattern can be gathered and analyzed: seasonal patterns of
movement, diurnal movement habits, gap crossing occurrences, response in movement patterns to
simulated stimuli etc. With a better understanding of how songbirds move and how urbanization
effects songbird movement, it will allow future planners to make informed decisions regarding
conservation and urban layout.

24

LITERATURE CITED
Adriaensen, F., J. Chardon, G. De Blust, E. Swinnen, S. Villalba, H. Gulinck, and E. Matthysen
(2003) The application of “least-cost” modelling as a functional landscape model. Landscape and
urban planning 64: 233–247.
Andrén, H. (1994) Effects of habitat fragmentation on birds and mammals in landscapes with
different proportions of suitable habitat: a review. Oikos 71: 355-366.
Betts, M. G., G. J. Forbes, A. W. Diamond, and P. D. Taylor (2006) Independent effects of
fragmentation on forest songbirds: an organism-based approach. Ecological Applications 16:
1076–1089.
Bonter, D. N., and E. S. Bridge (2011) Applications of radio frequency identification (RFID) in
ornithological research: a review. Journal of Field Ornithology 82: 1-10.
Brewer, L. W., C. A. Redmond, J. M. Stafford, and G. E. Hatch (2011) Marking ruby‐throated
hummingbirds with radio frequency identification tags. The Journal of Wildlife Management 75:
1664–1667.
Bridge, E. S., and D. N. Bonter (2011) A low‐cost radio frequency identification device for
ornithological research. Journal of Field Ornithology 82: 52–59.
Bridge, E. S., K. Thorup, M. S. Bowlin, P. B. Chilson, R. H. Diehl, R. W. Fléron, P. Hartl, R.
Kays, J. F. Kelly, W. D. Robinson, and M. Wikelski (2011) Technology on the move: recent and
forthcoming innovations for tracking migratory birds. BioScience 61: 689–698.
Brown, M. B., and C. R. Brown (1988) Access to winter food resources by bright-versus dullcolored house finches. Condor 729–731.
Collinge, S. K. (1996) Ecological consequences of habitat fragmentation: implications for
landscape architecture and planning. Landscape and urban planning 36: 59–77.
Desrochers, A., and S. J. Hannon (1997) Gap crossing decisions by forest songbirds during the
post‐fledging period. Conservation Biology 11: 1204–1210.
ESRI. 2015. ArcGIS Desktop: Release 10.3. Redlands, CA: Environmental Systems Research
Institute.
Fahrig, L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology,
Evolution & Systematics 34: 487–515.
Fahrig, L. (1998) When does fragmentation of breeding habitat affect population survival?
Ecological Modelling 105: 273–292.
Gobeil, J. F., and M. A. Villard (2002) Permeability of three boreal forest landscape types to bird
movements as determined from experimental translocations. Oikos 98: 447–458.

25

Graham, C. H. (2001) Factors influencing movement patterns of Keel-billed Toucans in a
fragmented tropical landscape in southern Mexico. Conservation Biology 15: 1789–1798.
Hansen, M. C., S. V. Stehman, P. V. Potapov, and I. Y. Fung (2010) Quantification of global
gross forest cover loss. Proceedings of the National Academy of Sciences of the United States of
America 107: 8650-8655.
Hill, G. E., and R. Montgomerie (1994) Plumage colour signals nutritional condition in the house
finch. Proceedings of the Royal Society of London. Series B: Biological Sciences 258: 47–52.
Hokit, D. G., B. M. Stith, and L. C. Branch (1999) Effects of landscape structure in Florida
scrub: a population perspective. Ecological Applications 9: 124–134.
Hou, L., M. Verdirame, and K. C. Welch (2015) Automated tracking of wild hummingbird mass
and energetics over multiple time scales using radio frequency identification (RFID) technology.
Journal of Avian Biology 46: 1–8.
JMP, Version 12. SAS Institute Inc., Cary, NC, 1989-2016.
McGraw, K. J., and G. E. Hill (2000) Plumage brightness and breeding-season dominance in the
house finch: a negatively correlated handicap? The Condor 102: 456–461.
McRae, B. H., and P. Beier (2007) Circuit theory predicts gene flow in plant and animal
populations. Proceedings of the National Academy of Sciences 104: 19885–19890.
McRae, B. H., B. G. Dickson, T. H. Keitt, and V. B. Shah (2008) Using circuit theory to model
connectivity in ecology, evolution, and conservation. Ecology 89: 2712–2724.
Pither, J., and P. D. Taylor (1998) An experimental assessment of landscape connectivity. Oikos
83: 166–174.
R Core Team. 2016. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
Ricketts, T. H. (2001) The matrix matters: effective isolation in fragmented landscapes. The
American Naturalist 158: 87–99.
Robinson, S. K., F. R. Thompson III, T. M. Donovan, D. R. Whitehead, and J. Faaborg (1995)
Regional forest fragmentation and the nesting success of migratory birds. Science 267: 19871990.
Schneider, C. W., J. Tautz, B. Grünewald, and S. Fuchs (2012) RFID tracking of sublethal
effects of two neonicotinoid insecticides on the foraging behavior of Apis mellifera. PLoS ONE
7: e30023.
Shochat, E., S. B. Lerman, M. Katti, D. B. Lewis, and J. S. Brown (2004) Linking optimal
foraging behavior to bird community structure in an urban‐desert landscape: field experiments
with artificial food patches. The American Naturalist 164: 232–243.
26

Smith, M. J., G. J. Forbes, and M. G. Betts (2013) Landscape configuration influences gapcrossing decisions of northern flying squirrel (Glaucomys sabrinus). Biological Conservation
168: 176–183.
Sokolov, L. (2011) Modern telemetry: New possibilities in ornithology. Biology Bulletin 38:
885–904.
Taylor, P. D., L. Fahrig, K. Henein, and G. Merriam (1993) Connectivity is a vital element of
landscape structure. Oikos 68: 571–573.
Thompson, C. W., N. Hillgarth, M. Leu, and H. E. McClure (1997) High parasite load in house
finches (Carpodacus mexicanus) is correlated with reduced expression of a sexually selected
trait. American Naturalist 149: 270–294.
Tremblay, M. A., and C. C. St Clair (2009) Factors affecting the permeability of transportation
and riparian corridors to the movements of songbirds in an urban landscape. Journal of Applied
Ecology 46: 1314–1322.
Violino, B. 2005. RFID Business Applications. RFID Journal. Available from
http://www.rfidjournal.com/articles/view?1334 (accessed on May 21, 2016).
Yoder, J. M., E. A. Marschall, and D. A. Swanson (2004) The cost of dispersal: predation as a
function of movement and site familiarity in ruffed grouse. Behavioral Ecology 15: 469–476.
Zeller, K. A., K. McGarigal, and A. R. Whiteley (2012) Estimating landscape resistance to
movement: a review. Landscape Ecology 27: 777–797.

27