University of Southern California Dissertations and Theses

Identifying areas of high risk for avian mortality by performing a least accumulated-cost analysis

(USC Thesis Other)

Identifying areas of high risk for avian mortality by performing a least accumulated-cost analysis

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Identifying Areas of High Risk for Avian Mortality by Performing a Least
Accumulated-Cost Analysis

by

William Winters
The University of Southern California

A Thesis Presented to the
Faculty of the USC Graduate School
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
Master of Science
(Geographic Information Science and Technology)

December 2015

Copyright 2015 William Winters
ii
Table of Contents

List of Figures .................................................................................................................... iv
List of Tables ..................................................................................................................... vi
Acknowledgments............................................................................................................. vii
List of Abbreviations ....................................................................................................... viii
Abstract .............................................................................................................................. ix
Chapter 1 Introduction ........................................................................................................ 1
1.1 Motivation ................................................................................................................. 2
Chapter 2 Related Work...................................................................................................... 4
2.1 Migration .................................................................................................................. 4
2.2 Map Projection .......................................................................................................... 9
2.3 Least Accumulated-Cost ......................................................................................... 12
2.4 Species of Study ...................................................................................................... 19
2.4.1. Red-eyed Vireo ............................................................................................... 20
2.4.2. Kirtland’s Warbler .......................................................................................... 21
2.4.3. Golden-cheeked Warbler ................................................................................ 24
2.5 Avian Mortalities in North America ....................................................................... 25
Chapter 3 Methods ............................................................................................................ 30
3.1 Data Used ................................................................................................................ 30
3.1.2 Source and Destination Areas .......................................................................... 31
3.2 Development of the Resistance Raster ................................................................... 31
3.2.1. Calculation of Slope Resistance Values ......................................................... 32
3.2.2. Inclusion of Wind ........................................................................................... 34
3.3 Corridor Analysis .................................................................................................... 37
3.3.1 Application of Analysis to Other Species ........................................................ 39
iii
Chapter 4 Results .............................................................................................................. 43
4.1 Red-eyed Vireo ....................................................................................................... 43
4.1.1. Choice of Water Resistance Value ................................................................. 43
4.1.2. Wind resistance value ..................................................................................... 44
4.2 Kirtland’s Warbler .................................................................................................. 49
4.3 Golden-cheeked Warbler ........................................................................................ 55
Chapter 5 Discussion ........................................................................................................ 57
5.1 Kirtland’s Warbler .................................................................................................. 57
5.2 Golden-cheeked Warbler ........................................................................................ 61
5.3 Future Research ...................................................................................................... 63
References ......................................................................................................................... 67

iv
List of Figures
Figure 1 Maps of Red-eyed Vireo migration created from sighting location data. Source:
Lehigh University (left) and Cornell University (right).............................................. 6
Figure 2 The 12 regions of the Interrupted Goode Homolosine Projection. Source:
Steinwand (1994) ...................................................................................................... 10
Figure 3 Map of estimated migration pathway of Kirtland’s Warbler. Source: Bocetti et
al. 2014 ...................................................................................................................... 22
Figure 4 Records of Kirtland’s Warbler during migration up to 1972. Source: Clench
1973 ........................................................................................................................... 23
Figure 5 Spatial distribution of reputable spring (left) and fall (right) sightings of
Kirtland’s Warblers up to 2013. Source: Petrucha et al. 2013 .................................. 24
Figure 6 Depictions of the wind direction in the Gulf of Mexico for the spring (left) and
fall (right) seasons. Source: NOAA.gov……………………………………………35
Figure 7 eBird data depicting the higher concentrations of Red-eyed Vireo during the
spring (through the Florida panhandle; left) and fall (through peninsular Florida;
right) with the darker shades of purple symbolizing more sightings ........................ 39
Figure 8 eBird.org data depicting the locations of Kirtland’s Warbler sightings for the
spring (left) and fall (right) with the darker shades of purple symbolizing more
sightings Source: eBird.org ....................................................................................... 41
Figure 9 eBird data depicting locations of Golden-cheeked Warbler observations with the
darker shades of purple symbolizing more sightings Source: eBird.org .................. 42
Figure 10 Results of the sensitivity analysis on the water values without wind, Map B
depicts the flyway that most closely resembles that of the Red-eyed Vireo………..45
Figure 11 Predicted Migration corridors for Red-eyed Vireo without wind data for Fall
(left) and Spring (right) ............................................................................................. 46
Figure 12 Predictions of Red-eyed Vireo migration flyways using wind data for Spring
(left) and Fall (right) ................................................................................................. 47
Figure 13 Predicted migration flyway for Red-eyed Vireo using exaggerated cost values
for wind ..................................................................................................................... 49
Figure 14 Predicted migration flyway for Kirtland’s Warbler based on slope and water
resistance values only ............................................................................................... 50
v
Figure 15 Representations of the migration corridors of Kirtland’s Warbler for the spring
(left) and fall (right) created from a least accumulated-cost analysis that included
slope, water, and wind as influencing factors……………………...……………….52
Figure 16 Map of the predicted spring migration of the Kirtland’s Warbler. Recorded
sightings from Petrucha et al. (2013) are transposed on the map for validation of the
model……………………………………………………………………………….53
Figure 17 Map of the projected fall migration of the Kirtland’s Warbler created with a
least accumulated-cost analysis. Data of recorded sightings (Petrucha et al. 2013) are
layered on top of the map to validate the analytical method .................................... 54
Figure 18 Migration corridors for Golden-cheeked Warbler, Fall (left) and Spring
(right)………………………………………………………………………………..56

vi
List of Tables

Table 1 Resistance values assigned for wind direction .................................................... 36

vii
Acknowledgments
I thank all of the Spatial Sciences Institute faculty that helped make my experience at the
University of Southern California incomparable. I especially want to thank Dr. Travis
Longcore for his guidance and never ceasing encouragement with completing this thesis.
I also appreciate the assistance that my other committee members, Dr. John Wilson and
Dr. Karen Kemp gave me while working on my thesis.

My family and friends have been a remarkable resource for me and I thank everyone who
offered words of encouragement while I worked steadily towards completion.

viii
List of Abbreviations

AGL - Above Ground Level
AVHRR – Advanced Very High Resolution Radiometer
ANF_BCR – Atlantic Northern Forest Bird Conservation Region
BCC – Birds of Conservation Concern
BCR – Bird Conservation Region
BNA – Birds of North America
DEM – Digital Elevation Model
GIS – Geographic Information System
GPS – Global Positioning System
IDW – Inverse Distance Weighted
JCU – Jaguar Conservation Unit
NCDC – National Climatic Data Center
NOAA – National Oceanic and Atmospheric Administration
RSF – Resource Selection Function
UV – Ultraviolet

ix
Abstract
Millions of birds are killed every year during their annual migration by colliding with tall
communication towers and buildings. The goal of this study is to identify areas of
specific concern for avian species during migration by modeling potential migration
corridors for Red-eyed Vireo (Vireo olivaceus), Kirtland’s Warbler (Setophaga
kirtlandii), and Golden-cheeked Warbler (Setophaga chrysoparia) as a case study. These
avian species perform transcontinental migrations each year. This study uses a least
accumulated-cost analysis to predict probability of use of routes between winter and
summer ranges by analyzing the presumed energetic cost of changing altitude (in
response to topographic relief), traversing large bodies of water, and compensating for
wind. Previous descriptions of migration pathways depict straight lines that do not take
into account geographic barriers. This study compares the results of existing methods to
the least accumulative cost model. The completion of the analysis on Red-eyed Vireo
allows the same analysis to be performed on two more rare species, the Kirtland’s
Warbler and the Golden-cheeked Warbler. The results of this study show that least
accumulated cost analyses are a viable option to assisting in determining preferred
migration routes for migratory birds. Least accumulated-cost analyses demand significant
computing resources, which can prevent studies of this size from being performed.
Advances in technology now enable studies of this magnitude to be performed and this
study is a proof-of-concept to illustrate the potential benefits of integrating these analyses
into conservation planning.

Chapter 1 Introduction
Millions of birds are killed every year in collisions with tall communication towers and
tall buildings (Klem 1990, Longcore et al. 2012, Loss et al. 2014). Studies have examined
the effect on mortality rates of the characteristics of towers (Longcore 2006, Longcore
2008, Gehring et al. 2009, Gehring et al. 2011) and buildings (Klem 2009, Klem et al.
2009). The effects of landscape configuration on the number of birds colliding with
towers and buildings is still largely unknown. Avian mortality at obstructions is
influenced by the concentration of birds at an obstruction and the characteristics of the
obstruction. Concentration of Neotropical migrants at a location is likely influenced by
many factors, including the location of the breeding and wintering grounds, combined
with static landscape features at several scales (ridgelines, mountains, water bodies),
location of habitat patches that are somewhat fixed but vary over time, and ephemeral
phenomena such as day-to-day weather conditions and prevailing winds.
Three primary factors affect the flight paths of migrating birds at the local level:
(1) end destination, (2) fuel, and (3) wind (Alerstam 2011). The end destination is a long
term/distance measure that is a guide for the bird’s flight direction. Fuel is necessary for
birds to complete the migration. Stopping spots with foraging opportunities will influence
birds to alter off the shortest route. Wind plays an important role in the amount of energy
a bird must exert in order to arrive at the end destination. A tailwind is preferable because
it reduces the amount of energy required. The balance of these factors will result in the
optimal flight path for a migration.
2
1.1 Motivation
This project is not directed to relate the number of avian collisions with towers to
geographic features. It is, however, poised to reduce the amount of mortalities of
migratory birds. This is done by predicting migration corridors that depict where birds
concentrate during migration. This information has the ability to decrease avian mortality
by showing conservation organizations, governing bodies, and private enterprises the
areas where more concern should be given when constructing tall towers or in assessing
the risks of tall building construction. The development of this method of estimating
migration routes also allows researchers to focus on specific areas, locate potentially
important stopover points and better understand the risks facing species during migration.
The completion of this project will be valuable for the scientific community in
that it does show that least accumulated-cost analyses are able to be performed for birds
at this broad of scale. The previous methods of modeling bird migration do not take
geographic features into account as major influences on migration. We know barriers
have an effect on migration and the least cost analysis will help to model these effects.
This analysis compares the previous methods to the benefits of the least cost analysis.
This study is performed with the small migratory songbirds as the subject but the
concept of using the corridor tool at a continental scale can be used in many fields.
Waterfowl are another species that conduct long migrations perennially. Millions of
dollars are spent every year to aide in the conservation of waterfowl habitat. It is crucially
important that these resources are allocated appropriately. The use of a transcontinental
least accumulated-cost analysis could provide information as to the locations of high
concentrations of migrating waterfowl. This information could be used to ensure that
3
resources are being used correctly to conserve the habitats that waterfowl will visit during
their annual migration. This is just one example of how this study could be used to
inform future research directions.
This study is a proof of concept showing that least accumulated-cost analyses can
be used to accurately model bird migration flyways. When performing least cost analyses
it is critical that the variables are parameterized correctly. Results can be skewed easily
so care must be taken to ensure weights are assigned appropriately. This study used a
common species, the Red-eyed Vireo, of which the migration patterns are well
documented to determine how to correctly parameterize the variables. Once successful,
the same parameters were used to predict the migration flyways of two rarer species, the
Kirtland’s Warbler and the Golden-cheeked Warbler. The results of the analyses of the
rarer species show that least accumulated-cost analyses are capable of predicting bird
migration routes but also provides valuable information as to the attention to detail
necessary to create reliable results.
4
Chapter 2 Related Work
The annual migration performed by countless birds every year is an important part of the
global ecosystem and human development is making it increasingly difficult to complete.
This chapter begins with a review of bird migration and how it has been modelled
historically. A discussion of map projections is included because the different methods of
modeling migration require the projection to have certain attributes. A description of
similar analyses using the least cost method of modelling corridors is used to describe the
benefits of this type of analysis. This chapter concludes with a summary of the Red-eyed
Vireo and its migration, as well as the Kirtland’s Warbler and Golden-cheeked Warbler.
An examination of the effect human development is having on bird species will round out
the chapter, discussing the implications of creating a model to help identify migratory
routes.
2.1 Migration
The annual journey of migratory birds is an extraordinary example of the majesty
of the natural world. The feats that some of the world’s birds accomplish are
breathtaking. It is during this migration that human development takes its greatest toll on
the bird populations, from destruction of stopover habitats to the construction of deadly
barriers. This analysis can aide in the reduction of avian mortality by accurately
modelling migration patterns to provide information on conservation priority areas. This
involves analyzing the influential factors that affect migration.
The three most important factors in determining migration route are end
destination, fuel, and wind (Alerstam 2011). There are barriers such as mountain ranges
and large water bodies that have an impact on the preferred route of migration. Birds
5
allow a wind to veer them off course just to stay in a tailwind (Elkins 2004). The tailwind
allows the birds to save energy and migrating birds follow the tailwind up to the point at
which the extra mileage will begin to cost energy instead of saving energy. This analysis
obviously includes the end destination but does focus on geographic barriers and how
wind assists birds in overcoming them.
Large bodies of water can be significant barriers to migration. These can range in
size from the size of local bay to the Gulf of Mexico to the Pacific Ocean. The Black
Brant (Branta bernicla nigricans) is known to undertake a direct migration across the
Gulf of Alaska that stretches over 5000 km without a chance to refuel (Purcell and
Brodin 2007). They accomplish this from help of wind currents. A study of raptor
migration in southern Spain showed that most species were more inclined to cross the
Mediterranean Sea when there was a northern wind to aid the flight (Meyer et al. 1999).
This analysis of songbird migration contains two main barriers; the Appalachian
Mountains and the Gulf of Mexico. The gulf is significant because seasonal wind patterns
play an important role in whether vireo will cross or circumnavigate. A study in
Louisiana of correlations between wind direction and passerine movement showed that
the routes mimicked the wind direction (Able 1972). The same study analyzed wind
patterns and determined that only under specific conditions where a strong front entered
deep into the gulf were there preferable conditions for a southerly migration. Average
wind direction for the Gulf of Mexico is of a southeasterly direction for both spring and
fall .This influences passerines to circumnavigate during fall migration. The same wind
patterns persist during spring while they contrarily persuade birds to cross the gulf during
the spring migration. Many observers have reported underweight passerines along the
6
north gulf coast during spring migration (Moore et al. 1990, Kuenzi and Moore 1991).
This finding provides evidence that birds are crossing the gulf during the spring
migration even though the energy costs are high.

Figure 1 Maps of Red-eyed Vireo migration created from sighting location data. Source:
Lehigh University (left) and Cornell University (right)
Some birds wait at stopover sites until a favorable wind appears and then take off
in the preferred direction. A study of Turkey Vultures (Cathartes aura) in North America
showed that during fall migration crosswinds were used by the vultures to gain elevation
on the Appalachian Mountains, following which they would glide back down in a
southerly direction (Mandel et al. 2011).
7
In the past, estimation of migratory routes was based on recorded sightings of a
species. These estimates can be considered fairly accurate because they reflect actual
sightings of birds during the migratory period. The problem is that migration maps of
common species are vague because the species have a vast migration corridor. The rarer
species often do not have enough recorded sightings to create a reasonable depiction of
the migratory pathway. Radio transmitters have been used in the past but are short range
and are used more often during breeding or wintering seasons (Nakamura et al. 2005).
Advances in technology are enabling researchers to more accurately model the migration
patterns of migratory birds. Large birds are able to wear GPS (Global Positioning
System) collars which allow researchers to track their movements during migration
(Mandel et al. 2011). Recently, the use of satellite transmitters have enabled birds as
small as 100 grams to be tracked during migration (Fiedler 2009).
There have been methods developed to identify migration routes for a variety of
animal species using a GIS (Geographic Information System) approach that do not take
species distribution data into account. These methods use geographic features to predict
the corridors animals use to traverse an area. Few studies have tried to use this
technology to model bird movements. A simple answer for why this is would be that
birds fly and terrain would therefore play a lesser role in determining their preferred
route. While it may be true that birds can fly over obstacles but to say that geographic
features do not affect avian migration would be false. This project suggests that
geographic features play a more important role than previously believed and that they
may be used to predict migration routes of bird species.
8
Historically, there have been two methods of theoretically modelling avian
migration without species sighting records; constant bearing and shortest distance
(Gudmundsson and Alerstam 1998). The constant bearing method assumes that birds
know the exact direction of their destination from any given point on the Earth. They
follow a constant bearing through their entire migration in order to reach their destination
as mariners would use the North Star to navigate the open ocean. This however is not the
shortest distance that one could travel between two points on the Earth. Gudmundsson
and Alerstam (1998) proposed that birds may have the intuitive knowledge of shortest
distance and will be constantly changing bearing to reduce the amount of ground they
must cover.
Red-eyed Vireo have wide breeding and wintering ranges so connecting the two
with migratory routes can be difficult. It is a common species so there are a lot of data on
recorded sighting with which to create maps. Previous research shows that the birds
migrate through Mexico or follow the Florida peninsula to reduce the amount of water
necessary to cross (Figure 1). This preference is supported by data of recorded sightings.
The problem is that the birds also cross the Gulf of Mexico directly (Able 1972).
Crawford (1980) further supported this by analyzing the number of migratory birds killed
along the northern gulf coast. The data used to create the existing maps use sighting data
which are primarily unavailable for vast amounts of water bodies. This eliminates the
possibility of including bird sighting over the Gulf of Mexico.
The United States Fish and Wildlife Service (2014) describes the Kirtland’s
Warbler migration as following a narrow band directly between Michigan and the
Bahamas. They note that the migration pattern is most likely the same for both spring and
9
fall migrations. There are very few acceptable records of Kirtland’s Warblers being seen
during migration. These records do not however line up with the assumed notion that the
birds follow a straight line between wintering and breeding grounds. Most publications
do not attempt to create a migration map because of the limited amount of available data.
2.2 Map Projection
The coordinate system chosen to depict an area can influence results of analyses
performed on the study area. The choice of map projection for a study with a small study
area and a large-scale can have little effect on the outcome of analyses (Steinwand et al.
1995). With the increase in study area, it becomes more important to decide on the
correct map projection. The analysis in this thesis of the migration of Neotropical
passerines spans across both North and South America. In a geographic coordinate
system the cell size of raster data is represented in degrees. The size of a degree changes
with latitude. This creates error in spatial analysis because the cell size is not uniform
throughout the study area.
To solve this problem the data must be projected into a planar coordinate system
where the cell size remains constant for every cell in the study area. This transfer of data
from a geographic coordinate system to a plane coordinate system can cause significant
error if the wrong projection is selected or incorrect parameters are set (Usery and Seong
2001). Many different projections are available to choose from and historically there has
been a lack of guidance in selection of map projections.
In 1923, J. P. Goode developed the Interrupted Goode Homolosine map
projection. He recognized a solution to the downfalls of the standard projections in that
each had areas of high distortion. He decided to combine the Sinusoidal projection with
10
the Mollweide projection to limit distortion at a global scale. The Sinusoidal project has
little error near the equator but as it approaches the poles distortion increases. The
Mollweide project does the opposite as it has little error near the poles. The Interrupted
Goode Homolosine projection takes advantage of both projections by splitting the
projection at 40º44’11.8” North and South (Goode 1925). This latitude is where the two
projections match up making it the perfect spot to merge. The projection is divided into
12 sections of accurate depictions of landmasses. The division of the projection pushes
the distortions into the oceans so that analyses can be performed accurately on the land
masses (Figure 2).

Figure 2 The 12 regions of the Interrupted Goode Homolosine Projection. Source:
Steinwand (1994)
11
Steinwand et al. (1995) evaluated six map projections for use during analysis of
global datasets. They suggested that the Interrupted Goode Homolosine map projection
was best for global datasets because it resulted in the least amount of distortion across the
globe. The Interrupted Goode Homolosine map projection was chosen for the Global
Land Advanced Very High Resolution Radiometer (AVHRR) 1 km dataset as well as the
AVHRR Pathfinder 8 km dataset (Steinwand 1994). The resolution of raster data also has
an effect on accuracy. Usery and Seong (2001) quantified the error resulting from
resampling. They determined that there is minimal error at resolutions under 8 km. The
amount of distortion begins to increase steadily after 8 km.
While an equal-area map projection such as the Interrupted Goode Homolosine
provides for accuracy for least cost analysis it does not maintain equal distances or
bearings. This analysis compares the ideas that birds maintain constant bearing or follow
the shortest route during migration with the concept that there are more factors involved
than just the end destination. An orthodrome is the shortest distance between two points
on Earth. A loxodrome is longer than an orthodrome but it is easier to use in navigation
because it follows a constant bearing. Thus a second projection is needed. The Mercator
projection is preferred for the constant bearing because it was created for just that
purpose in maritime navigation. Determining the shortest distance between two points on
the globe can be achieved with azimuthal projections (Gudmundsson and Alerstam
1998). The method used to predict the migration patterns directly influences the map
projection used in the analysis.
12
2.3 Least Accumulated-Cost
An important focus in conservation is to locate corridors that link fragmented
habitats. Least cost path analyses are a common tool used to do this. They use a raster
with cost values assigned to each cell to determine the easiest path through a landscape.
This method provides a single cell wide path to the destination. The least accumulated-
cost method takes into account the cost values in addition to the distance (Etherington
2013). It can be used to identify a least cost corridor, which is more helpful when
analyzing bird migration. A corridor allows for variation in migratory routes that are
common with avian species.
In this thesis, I propose using a least accumulated-cost analysis to identify
preferred migration corridors that birds would fly across the continents of North and
South America. Red-eyed Vireo (Vireo olivaceus), which has a wide range spanning two
continents and a known migratory pattern, is used to develop a methodology that is
replicated for two rarer species. The methodology is used to examine the migration
patterns of the Kirtland’s Warbler and the Golden-cheeked Warbler. Both species are rare
and there is relatively little evidence on which to describe a migration corridor.
The method takes into consideration three main factors that would affect the flight
pattern; slope, wind, and large bodies of water (Alerstam 2011). Slope affects the flight
patterns similarly to elevation but is needed because not all locations with high elevations
provide a barrier to migration. It depends on the elevation from which the bird is
approaching. Not all birds refuse to fly over large water bodies. Some species prefer to
fly across water and will only fly over land when necessary. These are more the
exception than the rule. Migratory passerines like the Red-eyed Vireo do not prefer long
13
water passages. In this study, large water bodies are assigned high cost values, which
incorporates the difficulty of crossing water bodies into the corridor model. Wind had
have positive and negative values that represent whether the wind assists or hinders the
energy costs of migration.
The method used to perform this analysis began with calculating the cost distance
from both summer breeding grounds and wintering locations. Calculating the
accumulated-cost of traveling from one cell to a source location does this. Cost distance
was calculated for every cell in the study area for both the breeding and wintering habitat
locations. This study did not identify the least cost path because it is too specific and
unrealistic of migrating birds. The goal was to find the corridor in the landscape that
provides the easiest migration possible. A cost corridor displays the total cost to travel
between two locations for every cell in a raster. A cost path simply shows the easiest
single cell wide route between two locations (Etherington 2013). The vagueness of a
corridor allows for the variability of migration patterns. The need for refueling at
stopover sites and the advantageous use of a tailwind that might drift a bird off course
create this variability in the flight paths. Once cost distance rasters are created for both
habitat locations the least accumulated cost surface can be made. This surface is a
combination of the three cost distance rasters. For every cell in the study area, the least
cost path is found to both locations. The output raster depicts the area connecting the two
habitat locations that has the least cost to travel. The result is a clearly defined corridor
that represents the easiest migration route.
The least cost approach has been used to model portions of avian migration
pathways. Downs and Horner (2008) used least cost modelling to determine the most
14
efficient stopover locations for migrating birds. They examined a section of the migration
and ran least cost analyses between known stopover locations to determine the best route
for the birds. This shows that least cost analyses can be used for avian species.
It would be a rare instance that an animal would take the exact least cost path
through a landscape. It follows the path of least resistance but there are many variables
that cannot be taken into account. There could be temporary blocks to the least cost path
making the use of corridors ideal. This idea of using wider corridors allowing for
variability was tested by Cushman et al. (2009) in a study on Black bears (Ursus
americanus). He stated that a broader scale of study will include this needed variability.
His study consisted of 160 source locations and 160 destinations. He created single cell
least cost paths between all source and destination locations. He then summed all paths
together to show all possible routes. His study used the gene flow of bears in the north
portion of the study area to determine the ease of movement through the landscape. This
allowed the study to use empirical evidence to assign cost values instead of just expert
opinion. Cushman et al. (2009) were not able to perform validation of his results because
it would take extensive research of tracking many individual bears.
Having realized the importance of connectivity between habitats, Richard Walker
and Lance Craighead (1997) conducted an analysis linking the Greater Yellowstone,
Northern Continental Divide and Salmon-Selway ecosystems. They performed the
analysis three times, once each for elk (Cervus canadensis), grizzly bears (Ursus arctos)
and cougars (Puma concolor). They were required to assign different weights to the
landscape inputs for the cost surface. The three inputs were habitat quality, amount of
forest, and road density. The similarity in results shows limited sensitivity based on the
15
weights of inputs to the cost surface. Sensitivity to changes in resistance values can vary
between species. The species chosen for this study have similar habitat preferences so it
makes sense that they would have similar migration corridors.
While studying Desert bighorn sheep (Ovis canadensis nelson) it was shown that
different species have their own methods and preferences of traversing landscapes (Epps
et al. 2007). These varying preferences can make predicting corridors difficult. It is often
the case that expert opinions are used to determine the weighting structure of cost
surfaces in corridor analyses. These expert opinions can cause error in prediction models
because they could vary based on who the expert is. It is important that empirical
evidence is used to assign cost values. This is often difficult if not impossible because of
the lack of empirical evidence of movement preference. Epps et al. (2007) used genetic
flow data collected from multiple sheep populations to provide this evidence. After the
study, tracking collars proved the model successful.
Sawyer et al. (2011) expanded on the work of Epps et al. (2007) by comparing it
to 23 similar studies. Their literature review resulted in the identification of three main
faults exhibited in least cost studies. The first is that studies are performed solely on
remotely sensed data without consideration as to whether or not habitat data is a good
measure of ease of movement. Second is the lack of empirical evidence and use of expert
opinion. They state that studies “must clarify biological processes on which resistance
values are based.” Whether expert opinion or empirical evidence is used to determine
cost structure it is important to perform a sensitivity analysis to see the degree to which
variance in cost structure will affect results. Thirdly, there must be validation of results.
Least cost path analyses can be very useful to conservation in that they show areas that
16
link fragmented habitat. The multiple causes of error show that they must only be used as
a first step in the conservation plan and must be validated by field studies. Sawyer et al.
(2011) showed that only nine of 24 studies were able to validate results.
The scale to which resistance values are assigned to cost surfaces can greatly
affect the results of a least cost path analysis. Beier et al. (2009) performed a theoretical
study on the effects of varying cost values. They studied eight species using four habitat
factors and showed little difference between models created using varying cost structures
for the five herbivore species. The three carnivorous species had alternate results showing
significant difference based on various cost allocations. Similarly, a different study
showed that the assignment of cost values made a significant difference in the results
(Dreizen et al. 2007). Their study of hedgehogs (Erinaceous europaeus) showed that in
unfamiliar territory it is possible to predict movement when the proper resistance values
are applied. A study of the Florida panther (Puma concolor) used a least cost path
analysis to determine possible routes for panthers to use to perform northern migrations
as their existing habitat is being destroyed (Kautz et al. 2006). This study performed a
sensitivity analysis and determined that the weights of cost values made little difference
in results. These examples show that the results of least cost analyses are dependent on
many variables including but not limited to species, habitat, landscape, and scale.
Brooker et al. (1999) created a dispersal model to determine if birds use landscape
corridors in flight. The birds studied were Blue-breasted Fairy-wrens (Malurus
pulcherrimus) and White-browed Babblers (Pomatostomus superciliosus). Their analysis
did conclude that birds are more likely to traverse through corridors of preferred habitat.
17
These are sedentary birds, meaning the study only demonstrates that non-migratory birds
move through preferred habitat at small scales.
A study of the Speckled Wood Butterfly (Pararge aegeria L.) (Chardon et al.
2003) examined the difference between cost and Euclidean distances. Researchers
recorded the locations of butterfly sightings along paths in two two-hectare sections of
forest in Belgium. A cost distance surface was created to represent a theoretical model of
movement through the study area. The presence-absence data collected in the surveys
provided a control to compare both the cost distance and Euclidean distance surfaces. The
results showed that the cost distance was a much better predictor of movement through
the study area. Although the Euclidean distance may be the shortest route is may often
not be the easiest route for an animal to use.
In the wake of the Wenchuan earthquake in 2008, important Giant Panda
(Ailuropoda melanoleuca) habitats in the Wolong Nature Reserve in China were found to
be destroyed or isolated. Li et al. (2010) conducted a least cost path analysis to determine
the best corridors between the fragmented panda habitat. The results provided
information on how and where the reconstruction efforts should be performed. Meegan
and Maehr (2002) used land cover data to perform a least cost path analysis to identify
corridors for migration of the Florida Panther to new potential habitats north of the
Caloosahatchee River. Forest, urban and roads layers were used in the analysis with
buffers applied to each. The analysis successfully provided a path to the new potential
habitat. The results were validated with radio telemetry data. Between the years 1998 and
2000 three radio collared panthers were recorded crossing the river within four km of
where the least cost path analysis predicted.
18
The California Tiger Salamander (Ambystoma californiense) is a threatened
amphibian species which uses specific habitats. Wang et al. (2009) created a model of
corridors between preferred salamander habitats using a least cost path analysis.
Researchers surveyed salamanders in 12 ponds to perform a gene flow analysis. A least
cost path analysis was then conducted between each of the ponds. The habitat was
classified as grassland, chaparral, and oak woodland. The least cost path analysis in this
study was used to determine the cost of moving through habitat instead of creating the
corridors.
The use of expert opinions to create a cost raster can increase the risk of error in
the results of a least cost path analysis. Opinions are inherently subjective and depending
on the expert providing the opinion it can alter the outcome of the study. Chetkiewicz and
Boyce (2009) developed a model that uses resource selection functions (RSF) to develop
cost rasters to be used in least cost path analyses. The RSFs were created from telemetry
data of cougars and grizzly bears wearing GPS collars. The telemetry data provided a
model of preferred habitat from which they created the RSFs. High values of RSF meant
preferred habitat so they were inverted to create the cost raster. The results showed that
the RSFs can be used to more accurately model least cost paths than expert opinion.
Until recent years, most least-cost analyses created single cell wide paths between
source locations. With the addition of the corridor function in ArcMap it is possible to
model more realistic corridors of wildlife migration. Rabinowitz and Zeller (2011) used
the corridor function to perform a least accumulated-cost analysis modeling linkages
between Jaguar Conservation Units (JCU) in Central and South America. The corridor
function then calculates a least-cost path for every cell in the study area between the two
19
source locations that passes through the cell. These paths are all summed together to
create a model of least-cost corridor.
The landscape characteristics used in Rabinowitz and Zeller’s study were land
cover, percentage tree and shrub cover, elevation, distance from roads, distance from
settlements and human population density. All layers were resampled to one km
2
and
projected in the same planar coordinate system. This study averaged the opinions of 15
experts who assigned cost values between 0–10. The six rasters were summed together
with the raster calculator function. The cost distance function was run for each of the 90
JCUs included in the study. The corridor function was used for all proximate pairs of
JCUs. The researchers used the minimum mosaic method to combine all the corridors
together. All corridors found that had a width less than 10 km were marked as a concern
that they may be severed if no actions were taken. The results of the study were validated
with presence/absence data. This analysis on the Red-eyed Vireo followed a similar
methodology to determine corridors of high bird concentrations during annual migration.
2.4 Species of Study
This study was aimed to test the possibility of using a least accumulated-cost
analysis to model bird migration patterns. It was decided to use one common species,
Red-eyed Vireo, of which the migration patterns are known to determine how to set the
parameters of the study. Once the parameters were set so that the results of the analysis
resembled the actual migration flyway the same parameters were transferred to two rarer
bird species, the Kirtland’s Warbler and Golden-cheeked Warbler. The following sections
provide background information about these bird species that were included in the study.

20
2.4.1. Red-eyed Vireo
The Red-eyed Vireo is a small songbird that is one of the most common species in
eastern North American forests. Red-eyed Vireo undertake a transcontinental migration
every year. The peak periods of migration for the Red-eyed Vireo are in April and
September. The Red-eyed Vireo have an extensive breeding range spanning from British
Columbia, Canada to the northern part of Florida. Saskatchewan, Manitoba, Ontario,
Quebec, Newfoundland are the Canadian provinces that hold vireo in the breeding
season. The United States of America have vireo spending the breeding season in the
Midwest, the northeast, the southeast, and the states of Washington, Oregon, Idaho and
Montana. The annual migration takes the Red-eyed Vireo down to South America.
Colombia, Venezuela, Guyana, Suriname, Ecuador, Peru, and western Brazil are the
recorded wintering ground for the Red-eyed Vireo (Cimprich et al. 2002).
A migration from North to South America would either need to follow through
Mexico, down the peninsula of Florida or cross the Gulf of Mexico. There is evidence
that suggests that some Red-eyed Vireo do cross the Gulf of Mexico during their annual
migration (Crawford 1980). A number of variables, however, factor in to whether the bird
will attempt the crossing. Favorable weather and the amount of fat storage are the two
that researchers have agreed upon (Able 1972, Alerstam 2011). Poor weather conditions
may persuade a bird to utilize the safer route around the gulf. Records of captured birds
during the spring migration on north coast of the Gulf of Mexico showed a mean mass of
15.9 grams which is only 1.3 grams above the fat free body mass of Red-eyed Vireo
(Cimprich et al. 2002). Captured birds at the same location during the fall migration,
before they cross the gulf, showed a mean mass of 19.7 grams, 5.1 grams above fat free
body mass. This finding indicates that a large amount of fat storage is needed to cross the
21
Gulf of Mexico. Mean mass of vireo during the beginning stages of the fall migration is
only 20.7 grams for an adult male (Cimprich et al. 2002). The implication of this
measurement is that only a single gram is lost as the bird migrates across North America
but it loses four grams crossing the gulf.
2.4.2. Kirtland’s Warbler
The Kirtland’s Warbler (Setophaga kirtlandii) is one of the rarest songbirds in
North America (Bocetti et al. 2014). They have a very specific preference to habitat for
both breeding and wintering. The breeding area for the species is mostly located in the
northern section of the lower peninsula of Michigan. There have been records of some
nests expanding into eastern Wisconsin. The Kirtland’s Warbler winters almost
exclusively on a single island in the Bahamas, Eleuthera. Existing maps of Kirtland’s
Warblers depict a straight line from Michigan to the Bahamas (Figure 3).
22

Figure 3 Map of estimated migration pathway of Kirtland’s Warbler. Source: Bocetti et
al. 2014
Estimates of the species population fell to as low as 400 individuals in 1972.
(Mayfield 1972). The reduction of the preferred habitat was critical to the decline in
population. Conservation efforts have been effective and by 2012 it is estimated that the
population is over 4000 individuals. Even though the population is growing there are still
threats to the species survival. Very few sightings of the Kirtland’s Warbler during the
migration months have been verified (Figure 4).
23

Figure 4 Records of Kirtland’s Warbler during migration up to 1972. Source: Clench
1973
Petrucha et al. (2013) analyzed records of previous sightings and showed that the
migration flyway of the Kirtland’s Warbler is not as narrow as it was previously
understood. Records of sightings range across the entire eastern United States (Figure 5).
It is no longer reasonable to believe that an entire species will follow a narrow migration
flyway, even when the population is as small as the Kirtland’s Warbler. It has been noted
24
that the breeding range is expanding into Wisconsin and with this expansion the
migration flyway should expand westward as well.

Figure 5 Spatial distribution of reputable spring (left) and fall (right) sightings of
Kirtland’s Warblers up to 2013. Source: Petrucha et al. 2013
2.4.3. Golden-cheeked Warbler
The Golden-cheeked Warbler (Setophaga chrysoparia) is another rare and
endangered species that is at risk from collisions during its migration. The have been
records of the species at elevations over 1000 meters along the Sierra Madre Oriental
(Groce et al. 2010). The breeding range is located in central Texas while the wintering
range is in southern Mexico and spans into neighboring Central American countries
(Ladd and Gass 1999). The rarity of this species lends to gaps in knowledge pertaining to
habitat and behavior.
25
2.5 Avian Mortalities in North America
Millions of birds are killed by tall structures every year. There have been studies
that have evaluated the causes of these fatalities (Gehring 2009, Gehring 2011, Longcore
2006, Klem 2006). The studies can be divided into two categories: those that focus on
fatalities by collision with communication towers and collisions with buildings. The
characteristics of the two types of structures that affect mortality rates vary greatly.
Communication towers can be deadly obstacles to migrating birds. Gehring et al.
(2009) performed a study that examined the characteristics of these towers to determine
why more birds die at certain towers than at others. The study analyzed the number of
fatalities that occurred at 24 towers throughout the state of Michigan during peak
migration periods over three years. Three of the towers were over 147 meters above
ground level (AGL) while the remaining were under 146 AGL. The study showed that
towers equipped with flashing lights as well as non-flashing lights were responsible for
most of the fatalities at communication towers. It is predicted that fatalities could be
reduced by 50–70% by removing non-flashing lights from communication towers. The
study also showed a higher mortality rate at the taller towers but there were not uniform
light specifications so exact numbers could not be determined.
Gehring et al. (2011) used the same data that was collected for their previous
study to analyze the affect that guy wires have on mortality rates. It is noted that Red-
eyed Vireos were the most common species found during all study periods. The study
states that the birds are colliding with guy wires and that 69–100% of fatalities could be
prevented by constructing towers without guy wires. It is also shown that 68–86% fewer
26
fatalities occur at the shorter communication towers (note that the tall towers are
equipped with non-flashing lights).
Collisions with buildings occur at a higher rate than with communication towers
simply because there are an exponentially larger amount of buildings than towers. This
leads to the analysis of characteristics of buildings with higher mortality rates. In 1990,
Dr. Daniel Klem Jr. performed a study to show that glass windows do kill birds and how
to prevent avian fatalities. His study contained three experiments. One studied the
collision occurrence rate at two homes in Illinois. One in a rural setting and the other in a
suburban neighborhood. The study showed similar number of collisions and a mortality
rate of over 50%. The field experiment used five glass panes located on the edge of an
agricultural field with four of the five panes obstructed by bird deterrents. These
deterrents were silhouettes of a swooping flacon and an owl, wind chimes and a border of
lights around the glass pane. The results from the experiment showed that the deterrents
were not successful. The final experiment was with captured Dark-eyed Juncos in flight
cages. The flight cages were set up with two routes out; one clear and one obstructed. The
only obstructions that worked were a solid cloth drapery and very tight patterns. The
study cited an instance where an Indigo Bunting (Passerina cyanea) was banded after
being wounded by a collision with a window and the same bird was recovered dead after
hitting the same window one year later.
Dr. Klem stated in his 2009 study that the unintentional killing of birds by
windows is the largest human associated source of avian mortality except habitat
destruction. This later study follows a similar model to his 1990 study in that he used the
same field and flight cage experiments. This study provides ways to prevent collision
27
with windows by using Ultraviolet (UV)-reflecting and UV-absorbing window coverings.
These coverings offer no obstruction to humans looking out but portray obstructions to
birds, preventing collisions. The experiment with glass panes in the field showed that the
films successfully deterred birds from colliding with the glass panes. The flight cage
experiment showed that stripes of the UV-reflecting film and the ceramic frit dot pattern
were most successful in the deterrence of birds. It is cited that the Swarthmore College
has experienced as few as two collisions per year since installing windows with the
ceramic frit dots. Similarly Muhlenberg College installed the dots on one side of a
building and not a single collision had occurred in the first year since installation while
the other side of the building, with clear glass panes, saw 12 instances in that same year.
A study was performed on Manhattan Island, NYC in 2009 to determine the effect
the building design and landscape context have on the occurrence of birds colliding with
buildings (Klem et al. 2009). The grounds surrounding 73 buildings in Manhattan were
monitored daily over 56–58 days in the fall of 2006 and spring of 2007 to record deaths
of birds by colliding with the buildings. Each building was assigned values of percentage
of vegetation around the building and percentage of building façade made of glass. The
amount of vegetation is important and can lead to higher collision rates because it
provides habitat and also can be reflected in glass façade encouraging collisions. The
results of the monitoring showed that 475 birds were killed in the fall while only 74 were
killed in the spring. Over 80% of collisions were fatal during both study periods. Less
than three-percent of fatalities occurred at buildings with little reflective glass on the
façade of the building.
28
It is difficult to estimate the amount of birds that are killed every year by
communication towers. The most recent study was performed by Longcore et al. (2012,
2013). They were the first to incorporate variation in mortality by region and species. The
goal was to determine biological significance of the mortality rates of individual species.
Longcore et al. define a biological significant impact as one that would “adversely affect
a species or its habitat and could be expected to affect population growth or stability of
species and influence population’s long-term viability.” It is estimated that 6.8 million
birds are killed every year by collisions with communication towers (Longcore et al.
2012). The number of each species killed by towers was determined by multiplying the
estimate of mortality by region by the average proportion each species found in tower
kills by region. This study used many different reports of fatalities which increased the
accuracy compared to previous studies. Weighting studies by species number only within
the Bird Conservation Regions decreases the amount of geographic bias based on species
distribution. The results of the follow-up study showed that 97.4% of birds killed at
communication towers were Passerines (Longcore et al. 2013). 13 of 20 bird species
killed most frequently, by percentage of population, are identified as Birds of
Conservation Concern (BCC) or endangered. Every year, an estimated 9 and 5.6% of the
Yellow Rail and Pied-billed Grebe populations are killed by communication towers. Both
are on the BCC list.
Contrary to Klem (2009), Loss et al. (2014) stated the number of fatalities caused
by collisions with buildings is second to predation by feral and pet cats. The difference in
the statistics is because of an update in the estimate of number of birds killed by cats.
Loss et al. (2014) conducted a literature review to estimate the number of birds killed
29
annually by collisions with buildings. They reviewed 23 studies and collected over
92,000 fatality records. The study estimated between 365 and 988 million birds are killed
every year by colliding with buildings. The study also attempted to determine species
vulnerability. They standardized datasets and summed the counts for each species. Then
the counts were regressed by log10(x+1) by population size. This resulted in seven
species listed on the BCC list as being disproportionately vulnerable to building
collisions.
The creation of a method that can predict the migratory pathways of birds could
assist in developing policies to prevent avian mortality. As of now, finding dead birds is a
useful tool in determining the migratory pathways. Geographic features play an integral
role in which routes birds use during their migration. Being able to assign weights to
different features will allow researchers to predict migration. Once a basic methodology
is proven to be practical it can be expanded upon and used for more broad and specific
applications.
30
Chapter 3 Methods
A study of the Red-eyed Vireo is used to create a method of predicting migratory
pathways. A least accumulated-cost analysis is used to model the migration of Red-eyed
Vireo by assigning weights to three of the main factors determining their route. The
analysis combines multiple criteria in a single raster and determines the easiest route
through the study area from both the breeding and wintering locations. The least
accumulated-cost analysis combines both layers to create a corridor showing the easiest
routes between both locations. Once a correct model of the vireo migration was created,
the same parameters were used to predict the migration flyways of two rarer species, the
Kirtland’s Warbler and the Golden-cheeked Warbler.
3.1 Data Used
A least accumulated-cost analysis requires many data that come from various sources. It
is important to combine all of these data properly. Some data were used purely as a guide
for the parameterization of the resistance values, particularly data taken from the crowd-
sourced website, ebird.com. This data shows the locations of sightings of species of birds
around the world.
3.1.1 Digital Elevation Model
The GTOPO30 dataset includes DEM data for the entire globe. Copies of the
GTOPO30 dataset for the continents of North and South America were downloaded from
ArcGIS Online for use in this analysis. They were then merged together using the Mosaic
function in ArcGIS and saved to the local hard drive. It was then projected into the
Interrupted Goode Homolosine map projection with a resulting cell size of 1318.218206
meters. As mentioned earlier, the amount of error in the shape and size of the landmasses
31
increases rapidly with cell sizes greater than eight kilometers, which is why the
GTOPO30 data was chosen.
3.1.2 Source and Destination Areas
Red-eyed Vireo are a common species and much is known about their migration
patterns. This assists in the development of a method to predict migratory routes because
we know what the results should look like. A shapefile of the Bird Conservation Regions
(BCR) in North America was downloaded from the United States Geological Survey
website (www.pwrc.usgs.gov/bba/). It contained all of the BCRs while only the
ANF_BCR was needed. The source features for the Cost Distance function are the
Atlantic National Forest BCR (ANF_BCR) and the country of Ecuador. These locations
were selected because they are known as breeding and wintering locations of Red-eyed
Vireo. The path between the two locations have several barriers to migration which help
in testing the hypothesis that we can model the migration without species distribution
data. After the shapefile was imported into ArcGIS and the ANF_BCR was selected the
feature was exported to a local database as its own shapefile. The shapefile for Ecuador
was downloaded from ArcGIS Online and also copied to the local database.
3.2 Development of the Resistance Raster
A least accumulated-cost analysis spanning the entire length of North and South
America would require higher processing capabilities than available so the study area was
clipped to the exact extent necessary. The analysis of the Golden-cheeked Warbler had a
small study area west of the other two species so a separate raster needed to be created.
To do this, shapefiles were created and rectangles drawn around the study areas. The Clip
function then used this outline to create new rasters of only the necessary extent. The cost
32
rasters for the analysis were based off of a slope raster created from the DEM. This slope
raster was modified to include water and wind values.
3.2.1. Calculation of Slope Resistance Values
Elevation is not in itself a good metric of the difficulty of moving through space.
If the bird is already at 1000 m elevation then a cell of 1005 m would not be particularly
difficult to cross. Therefore, the change in elevation, slope, was used as the resistance
factor. Slope represents the steepness of the terrain. Heading up or downhill is a longer
distance and therefore more energy expensive to birds. Positive and negative slope values
were treated the same for this analysis because passerines are not great gliders and cannot
take advantage of the downhill. Slope was calculated using the Slope function in ArcGIS.
It is necessary to ensure the units of measure are the same for elevation and the X and Y
cell lengths of the DEM. When projecting the DEM into the Interrupted Goode
Homolosine projection the units of measurement for cell size were changed from decimal
degrees into meters. This resulted in the cell size and elevation being measured in meters
so the slope was calculated properly.
Least cost analyses require a cost raster that includes resistance values that can be
added evenly to determine the cost of travel away from the source location. The raster
was resymbolized from a stretched color ramp into 9 classes using the natural breaks
method. The natural breaks classification method divides the data into classes that are
based on natural groups in the data distribution. Most of the study area has a very low
slope value so it was important when reclassifying to maintain the natural distribution
patterns inside the data. Then the 9 classes were used to reclassify slope raster into values
ranging from one to nine. A value of one is the lowest value and therefore the easiest to
33
travel through. This range of values was chosen because it is a compromise between
having too few value classes and the alternative of too many. It maintains simplicity for
the integration of additional variables.
3.2.2. Water Resistance Value
The areas of water were represented by “NoData” values in the slope raster. This
comes directly from the original DEM that only included data for land masses. All cells
not representing landmasses were automatically assigned a value of “NoData”. Small
bodies of water that would not show up on the raster were excluded automatically. If the
water body is not at least the size of a whole cell then it was not included in the study.
Birds are able to fly over water so these cells needed to be included in the analysis. To do
this the raster was reclassified seven more times. The “NoData” values in the slope raster
were reclassified separately into values ranging from two to eight creating seven
individual cost rasters. Only water bodies large enough to be represented as “NoData” in
the DEM were considered water bodies for this analysis. The analysis is a proof of
concept and more attention to determining which water bodies constitute a barrier should
be made in future analyses. The different values for water were assigned in order to
perform a sensitivity analysis on the resistance values. The effect that alternate values
have on the outcome of the analysis was evaluated to determine the appropriate values to
properly model the migration patterns of the Red-eyed Vireo. Knowing that the vireo
cross large bodies of water but only under certain conditions means that when evaluating
the values I was looking for a model that displays the migration as crossing water and
land.

34
3.2.2. Inclusion of Wind
Wind plays an integral role in determining migration patterns for any species of
migratory bird (Alerstam 2011). Therefore it was important that it be included in a least
accumulated-cost analysis of bird migration. The first step in including wind was done by
identifying the wind patterns over the Gulf of Mexico. More broad wind data were
included later in the study. During the spring migration there are favorable south east
winds through the Gulf of Mexico (Figure 6). Able (1972) recorded winds as unfavorable
for trans-gulf migration during fall months. A new shapefile was created and an outline
of the Gulf of Mexico was sketched as a new feature to represent the springtime wind
patterns. This new feature did not include the southern area of the gulf west of the
Yucatan peninsula because while the winds are favorable it is still preferable that birds
stay over land where they can rest. The area of the gulf east of the Yucatan peninsula was
also omitted because if the birds head east from the peninsula they would be flying into a
head wind which would have a negative effect on energy expenditure.
This new shapefile representing the Gulf of Mexico was turned into a raster with
the Polygon to Raster function in ArcGIS. The raster was set to have the same
dimensions and orientation as the resistance raster so that values in this raster could be
used to modify the values in the resistance raster. The value for the cells inside the Gulf
of Mexico wind raster were reclassified to a value that would, when added to the cost
raster with the Raster Calculator, result in a value of one which represents the lowest cost
to travel. This now depicts a lower cost to crossing the gulf from the Yucatan Peninsula
to the northern gulf coast. Cost distance rasters were then created for both the breeding
and wintering locations so that representations of migration could be made with the
35
Corridor function. These wind data were only included in the analysis of spring migration
for the Red-eyed Vireo.

Figure 6 Depictions of the wind direction in the Gulf of Mexico for the spring (left) and
fall (right) seasons. Source: NOAA.gov
The inclusion of the Gulf of Mexico wind data made it clear that wind should be
calculated and included for the entire study area. There have been many attempts at
recording prevailing wind data. Few of these data for prevailing winds have been
documented in a way that could be easily included in a least cost analysis. This analysis
is a proof of concept so the most sophisticated data available are not necessary. I obtained
the average wind direction for cities across the United States from the National Oceanic
and Atmospheric Association’s (NOAA) website (NCDC 1998). The data are an average
of records from the National Climatic Data Center (NCDC) dating from 1930 to 1996.
To include the City-based prevailing wind data in the least cost analysis it needed
to be in raster format. A series of points were created as a new shapefile with one point
for each city. Every point was assigned a value of 1–8 to represent the prevailing wind
direction (Table 1). The points and their values were then interpolated using the Inverse
36
Distance Weighted (IDW) function. IDW interpolation assigns values to unmeasured
locations based on the values of measured locations with closer locations given a greater
influence. This created a raster of wind direction for the study area. Resistance values for
wind direction were determined by assuming that in a south west migration a north east
wind would reduce cost while a south west wind would increase cost. The wind value is
meant to adjust the overall resistance value of each cell. Each classification, one through
eight, was then reclassified to corresponding values from Table 1. A raster for spring and
fall were each produced with inverse values.

Table 1 Resistance values assigned for wind direction

The new wind direction rasters were added, using the addition key in the Raster
Calculator, separately to the cost raster. Then the cost raster had to be reclassified once
again because negative and zero values existed. Least cost analyses cannot run with
negative or zero values. The negative and zero values were reclassified to hold a value of
one because it is the lowest acceptable value. As with other factors the wind values had to
be tested for sensitivity to the results. Higher values were assigned to a new wind
Value Wind
Direction
Fall
REVI
Spring
REVI
Fall
KW
Spring
KW
Sensitivity Analysis
REVI
1 N -1 1 -1 1 2
2 NE -2 2 -1 1 3
3 E 0 0 0 0 0
4 SE 1 -1 2 -2 -2
5 S 1 -1 1 -1 -2
6 SW 2 -2 1 -1 -3
7 W 0 0 0 0 0
8 NW -1 1 -2 2 2
37
direction raster. This was used to create an additional cost raster that were used to assess
the effect different wind values have on the results.
3.3 Corridor Analysis
Cost distance rasters show the cost necessary to travel through space to every
point in the study area from the source location. Cost distance rasters from both the
source and end locations are necessary to create models of migration corridors. The Cost
Distance function in ArcGIS was used to create cost distance rasters from all seven cost
rasters for both the ANF_BCR and Ecuador. This resulted in 14 cost distance rasters.
Each time the function was performed the processing extent was set to match the study
area outline to ensure the entire study area was processed. The output coordinate system
was always set to the same as the input cost raster to keep the raster in the Interrupted
Goode Homolosine projection.
Corridors were created from all seven sets of cost distance rasters to analyze the
effect the assigned resistance values had on the outcome. Corresponding cost distance
rasters were entered into the Corridor function in ArcGIS to create seven models of
migration corridors. This depicts a narrower corridor than the broad corridor the initial
symbology represented. These corridors were compared to evaluate which resistance
value for the water cells resulted in the best representation of the migration corridors
recorded by Able et al. (1972), Kuenzi et al. (1991), Moore et al. (1990) and eBird data
(2015).
Every factor that is included in the cost raster must be analyzed to determine if the
values assigned deliver results that match what we already know of the migration routes.
Two additional cost distance rasters were created to test the sensitivity of resistance
38
values for slope. The slope value was tested by creating new slope rasters where the
values were resymbolized using equal interval and standard deviation classifications. The
change in resistance values exaggerated the difference in the lower values in slope to
extenuate the less than dramatic topography in the southeastern US. This analysis was
performed after the appropriate value for water bodies was found.
The various corridor maps with different resistance values were compared to
studies on the migration routes of Red-eyed Vireo. The species has a wide range so it is
difficult to find a map of only the birds migrating from a specific region. Able et al.
(1972), Kuenzi et al. (1991), Moore et al. (1990) and eBird data provided information
that was used to evaluate the migratory flight patterns predicted in the models (Figure 7).
The best models were chosen by comparing them to what we know about the migration
patterns of Red-eyed Vireo. Migrating Red-eyed Vireos travel down the Florida
Peninsula in the fall and more likely to cross the gulf in the spring. The results of the
analysis were compared to this pattern to select the best parameterization of the migration
model.
39

Figure 7 eBird data depicting the higher concentrations of Red-eyed Vireo during the
spring (through the Florida panhandle; left) and fall (through peninsular Florida; right)
with the darker shades of purple symbolizing more sightings
3.3.1 Application of Analysis to Other Species
The analysis of Red-eyed Vireo was meant to determine how to parameterize the
different variables influencing bird migration routes. Once these parameters were set they
were transferred to more rare species, the Kirtland’s Warbler and the Golden-cheeked
Warbler. Less is known about the rarer species’ migration patterns so the least
accumulated-cost analyses may be able to provide some insight into the flyways of rarer
species.
3.3.1.1. Kirtland’s Warbler Analysis
The cost-surface values that resulted in the corridor routes that best matched
known migration pathways for the common and well-studied Red-eyed Vireo were then
used to develop predicted migration routes for Kirkland’s Warbler. The flyway for the
40
Kirtland’s Warbler (Setophaga kirtlandii) is still unknown but the compiling of sighting
records by Petrucha et al. (2013) provides some information on the migratory routes. The
core breeding grounds for the species are located in the northern Lower Michigan and the
wintering grounds mainly on the island of Eleuthera in the Bahamas. These habitats were
used as the source locations for the analysis and used in concert with the cost rasters
created during the Red-eyed Vireo analysis.
The cost raster with the water value set at 2 was chosen because previous
migration models (Bocetti et al. 2014) assumed that the warblers are not opposed to
traveling across water. There are many records of warblers in Ohio which suggest the
crossing of Lake Erie (Figure 8). The resulting cost distance rasters were used to
calculate the migration flyway with the Corridor function in ArcMap.
Kirtland’s Warblers migrate in north-west and south-east directions. This required
that additional rasters be created to represent how wind affects flight paths. They were
created by reclassifying the ones used in the Red-eyed Vireo analysis (Table 1) to reflect
the different overall migration direction. These rasters were added and the analysis was
performed as before.

41

Figure 8 eBird.org data depicting the locations of Kirtland’s Warbler sightings for the
spring (left) and fall (right) with the darker shades of purple symbolizing more sightings
Source: eBird.org
To evaluate the results of the analysis, I created point shapefiles for the spring and
fall sightings of Kirtland’s Warbler compiled by Petrucha et al. (2013). Locations for
reliable sightings of Kirtland’s Warblers were transferred manually from Petrucha et al.
(2013) and were used to verify the results of the study. These shapefiles were then
overlaid on top of the results from the least accumulated cost analysis.
3.3.1.2. Golden-cheeked Warbler Analysis
A smaller study area was created for the least cost analysis of the migration of the
Golden-cheeked Warbler. Shapefiles of the breeding locations in Texas and wintering
locations in Central America were created for use in the analysis. Similar processes for
42
the Red-eyed Vireo and Kirtland’s Warbler were used to perform the analysis on the
Golden-cheeked Warbler’s migration patterns.
Slope, water bodies and wind were all included in the analysis. As in the Red-
eyed Vireo analysis, a value of three was given to the water bodies. The analysis was
performed with and without the NOAA wind data. This is because the points that were
used to create the raster are only located in a small area of the study area and do not
reflect the local wind patterns. A separate raster was created to represent the inclusion of
the south-east winds that occur in the Gulf of Mexico, similar to that of the vireo analysis.
This raster was only included in the creation of a spring migration map.
Verifying the results of the Golden-cheeked Warbler analysis was difficult
because of the lack of information on the migratory route of the species. Only the
research of Groce et al. (2010), Ladd and Gass (1999) and the eBird data (Figure 9) could
be used to assess the results.

Figure 9 eBird data depicting locations of Golden-cheeked Warbler observations with the
darker shades of purple symbolizing more sightings Source: eBird.org
43
Chapter 4 Results
The least accumulated-cost analyses performed as part of this study provided a wide
array of results. The Red-eyed Vireo analysis provided information on how to correctly
assign weights to the different criteria involved in the analysis. The results of the vireo
analysis enabled the methodology to be transferred to rarer species. The Kirtland’s
Warbler analysis showed how that this technology can be used to accurately predict
migration pathways of birds. The analysis of the Golden-cheeked Warbler highlighted the
shortcomings of least accumulated-cost analyses by providing results that are not
supported by species sighting records. This is useful information because it emphasizes
the importance of including all necessary information and assigning the proper values to
factors specific to individual species that affect flight patterns.

4.1 Red-eyed Vireo
As mentioned before, the Red-eyed Vireo analysis was used to determine the
appropriate values for the different resistance values. This section details the outcome of
the Red-eyed Vireo analysis and describes how the values were chosen from the results
of the analysis.
4.1.1. Choice of Water Resistance Value
The review of existing migration flyway information for the Red-eyed Vireo
showed the preferred routes of migrating vireo. The results of the sensitivity analysis for
the resistance values of water bodies were analyzed to determine which value provided
the most accurate depiction of the flyways. Existing information about migrating vireo
shows that the birds do not prefer to fly over large bodies of water but will if the
44
conditions are right. A value of three for the water resistance value portrays this behavior.
A value of two would create a map that would have the vireo only fly over water which is
unlikely because there is not any sources of food to replenish necessary energy (Figure
10). Values higher than three would depict migration patterns that would show that vireo
would not fly over water for any reason, which is not true.
4.1.2. Wind resistance value
The addition of wind values to the analysis provided a level of detail that was
unattainable with only slope and water bodies. Including only the wind layer for the Gulf
of Mexico provided a spring migration where the vireo cross directly over the Gulf of
Mexico and has the fall migration as completely avoiding the gulf (Figure 11). The two
major pathways for vireos from the east coast are to follow the Florida peninsula down to
the Caribbean or perform a direct gulf crossing between the Yucatan peninsula and the
northern gulf coast. The inclusion of the wind layer in the cost raster depicted routes that
represent these notions. The fall migration is depicted as following Florida south and
crossing through the Caribbean or performing a trans-gulf flight directly to the Yucatan
Peninsula (Figure 12). The maps for the spring migration show that the vireo cross the
gulf from the Yucatan Peninsula and may follow along either the east or western side of
the Appalachian Mountains (Figure 12). That is, based on slope, water, and wind alone,
the model predicts a path through peninsular Florida in the fall and a path that bypasses
the peninsula and makes landfall in the panhandle in the spring.
45

Figure 10 Results of the sensitivity analysis on the water values without wind, Map B depicts the flyway that most closely resembles
that of the Red-eyed Vireo
46

Figure 11 Predicted Migration corridors for Red-eyed Vireo without wind data for Fall (left) and Spring (right)
47

Figure 12 Predictions of Red-eyed Vireo migration flyways using wind data for Spring (left) and Fall (right)
48

The inclusion of the wind values required another sensitivity analysis to take
place. The results of this analysis showed that the initial values set for wind in the cost
raster provided results that resembled patterns of migration that have been observed. The
higher values exaggerated the effect wind had on the outcome. The resulting map
depicted the flyway for the Red-eyed Vireo as traversing the Atlantic Ocean and only
occasionally crossing land (Figure 13). This is a near impossible route for a migratory
passerine to undertake and therefore provides the evidence needed to be assured that
reasonable wind values were used in the initial cost raster.
Figure 7 shows data collected from the crowdsourced database eBird that depicts
high concentrations of Red-eyed Vireo in peninsular Florida during September and
higher concentrations of the vireo along the northern coast of the gulf during the month
of April. This is further evidence supporting the claim that Red-eyed Vireo choose to
cross the gulf during spring migration but will circumnavigate during fall.

49

Figure 13 Predicted migration flyway for Red-eyed Vireo using exaggerated cost values
for wind
4.2 Kirtland’s Warbler
The least accumulated-cost analysis of the migration pattern for the Kirtland’s
Warbler resulted in a model (Figure 14) that does not mirror the model proposed by
Bocetti et al. (2014) in Birds of North America. The initial analysis, not including wind,
depicts a corridor that heads south from the breeding ground and includes all of Ohio and
Indiana (Figure 14). The model then splits as some birds venture across the southern
50
Appalachian Mountains through West Virginia. There is another split in the flyway in
Tennessee that shows a possible area for high concentrations of migrating warblers. The
widest section of the flyway continues south to the gulf coast before heading east to the
Bahamas.

Figure 14 Predicted migration flyway for Kirtland’s Warbler based on slope and water
resistance values only
51
The results of the analysis of spring migration showed only a slight difference
from the analysis without wind. That difference is that the warblers crossed Lake Erie
and ventured into Ontario before settling back in Michigan (Figure 15). The map of fall
migration with wind showed that the warblers began the migration by heading almost due
east from Michigan (Figure 15). The birds then proceeded south but stayed away from
the coast and flew along the eastern side of the Appalachian Mountains. They do not
cross over to the Bahamas until down in southern Florida. The map of spring migration
created in this analysis contained 256 (82%) of the 311 data points provided by Petrucha
et al. (2013) while the fall migration map contained 64 (56%) of the 114 data points
(Figures 16 & 17).
52

Figure 15 Representations of the migration corridors of Kirtland’s Warbler for the spring (left) and fall (right) created from a least
accumulated-cost analysis that included slope, water, and wind as influencing factors

53

Figure 16 Map of the predicted spring migration of the Kirtland’s Warbler. Recorded
sightings from Petrucha et al. (2013) are transposed on the map for validation of the
model
54

Figure 17 Map of the projected fall migration of the Kirtland’s Warbler created with a
least accumulated-cost analysis. Data of recorded sightings (Petrucha et al. 2013) are
layered on top of the map to validate the analytical method
55
4.3 Golden-cheeked Warbler
The least accumulated-cost analysis of the Golden-cheeked Warbler’s migration
predicted a migration corridor that follows the east edge of the Sierra Madre Oriental
mountain range and extends to the gulf coast of Mexico and Central America (Figure 18).
The analysis that included an adjustment for the prevailing southeastern wind in the Gulf
of Mexico offered different results. This model showed that it is possible that Golden-
cheeked Warblers undertake a trans-gulf flight (Figure 18). These prevailing winds are
only advantageous to migrating birds during a northern migration so it was only included
in the analysis of spring migration. The resistance value of water is enough to prevent the
analysis from predicting a gulf crossing during fall when the winds are not advantageous.
56

Figure 18 Migration corridors for Golden-cheeked Warbler, Fall (left) and Spring (right)
57

Chapter 5 Discussion
The analysis reported here is different than previous studies in that it did not use expert
opinion to determine the resistance values of the cost raster. Expert opinions were used
but only to determine the already known migration patterns of the Red-eyed Vireo. The
resistance values were determined through a series of sensitivity analyses that resulted in
a model that is able to create maps that depict an accurate representation of the known
migration patterns. The accurate depiction of the vireo migration shows that the
resistance values are set at the proper level and can be transferred to other species. It
should be noted that every species is different and some alterations to the resistance
values must be made. The results of this study are encouraging to the concept of
modeling long distance bird migration using a least accumulated-cost analysis. While this
study ignored important criteria involved in migration it was useful in proving that the
method is sound and could be expanded upon in future research.
5.1 Kirtland’s Warbler
The map of the migration flyway provided by the Birds of North America (BNA)
only stretches as wide as a single state with a due heading towards the Bahamas from the
Michigan breeding range. This map excludes Indiana and the states west of the
Appalachian Mountains except Ohio. Petrucha et al. (2013) collected records of
migratory Kirtland’s Warblers and included 425 records in a study to describe the bird’s
migratory patterns. In that study, Indiana and Illinois have more acceptable sightings than
West Virginia, Virginia, Pennsylvania and North Carolina. The latter are states that the
BNA map depicts as making up most of the corridor. There are substantial records in
58
these states but the migration corridor for the Kirtland’s Warbler extends outside of the
BNA model.
The findings of Petrucha et al. (2013) similarly show that there is a difference
between the flyways used during the spring and fall migrations (Figure 5). Prevailing
wind patterns have a substantial influence on the flight patterns of migrating birds. The
inclusion of wind into the analysis resulted in alternate flyways for spring and fall
migrations. These predicted migration flyways were compared to the data provided by
Petrucha et al. (2013). There will always be considerable variation in bird migration but
accuracy ratings of 82 percent for spring migrations and 56 percent for fall migrations
illustrates that a least accumulated-cost analysis is a promising method of predicting bird
migration flyways. While 56% does not appear to be a promising number it is promising
because some of the sightings outside of the predicted flyway are of birds that breed in
Wisconsin which would have them begin their migration west of the breeding location
used in this study.
The predicted model of the spring migration illustrates the birds cross over into
Florida directly before heading north to the breeding grounds in northern Michigan. It is
apparent that the warblers decide to cross the Appalachian Mountains as soon as possible
if they do not circumnavigate south of the barrier. The map shows a section of the
mountain range where birds do not want to cross but within that stretch there is a break in
the barrier which could be a critical migration corridor. This break in the barrier is
probably the result of the locations of the Pigeon and French Broad Rivers. These two
rivers cross the mountain range from North Carolina into Tennessee. The rivers have cut
out low lying areas that allow for easier crossing for the birds. The data points from
59
Petrucha et al. (2013) form a line that crosses through this area and there are eBird
records of sightings in the area. From that point the warblers are able to head north to the
breeding ground. Similar to Petrucha et al. (2013) and this analysis, the eBird data shows
more sightings west of the Appalachians and in the state directly south of Michigan. A
majority of the data points are inside Michigan, Ohio and Ontario because they are in
close proximity to the breeding grounds. This results in a high percentage within the
corridor boundary. Some birds head west of Lake Michigan and nest in northern
Wisconsin. This explains the data points bordering the southwestern end of the lake.
A dramatic difference is seen between the fall and spring migration maps. While
the spring has the birds mostly to the west of the Appalachians, the fall map shows a
more eastern route. The fall migration map predicts that the warblers head east from their
Michigan breeding grounds. They travel through Ontario and Ohio to reach Pennsylvania
before turning south. The spring and fall corridors share a border cutting West Virginia in
half while only sharing a margin of the state. Once across the mountains the warblers
then follow south along the side of the range, avoiding the coast. The eBird data include
more sightings of the Kirtland’s Warbler on the east side of the Appalachians during fall
migration further lending support to the model. There are few records of the warblers
being spotted along the coast until Georgia. The birds then continue south over land until
they are close enough to the Bahamas to traverse the sea.
The only large group of recorded fall sighting of the Kirtland’s Warbler that do
not align with the predicted fall migration flyway are those bordering the western shore
of Lake Michigan. This is easily explained by the spread of the Kirtland’s Warbler
breeding area into northern Wisconsin. This analysis did not include these breeding
60
grounds, which resulted in the lower than expected results in the assessment of model
accuracy. Inspection of the data points in Petrucha et al. (2013) indicate that it could be
possible that the warblers that nest there perform a different migratory pattern then their
Michigan neighbors. A separate analysis should be performed for the different breeding
areas instead of trying to combine them in future analyses.
The specificity of breeding habitat requirements for the Kirtland’s Warbler
presents a threat to the future survival of the species. While work is being done to
preserve critical habitat in the breeding range, little is known of the habitat the birds use
as stopover points. These stopover points are crucial to migrating birds as they provide
much need food to fuel the continuation of their journey. It is possible that the warblers
follow specific migration routes in order to make use of certain habitats as stopover
sights. This information should be included in future least cost analyses. Land cover data
describing the preferred habitat could be assimilated into the cost raster. This would show
that areas with preferred habitat are of lower energy cost because they actually increase
the physical energy of the birds. The inclusion of this information could lead to a more
accurate model of the migration routes of Kirtland’s Warblers.
Existing models of the migration flyway for the Kirtland’s Warbler depicts a
constant bearing between the species’ breeding and wintering locations. It was once
customary practice to believe that the end destination was the sole focus of the bird and
any detours were due to weather and purely accidental. This also implies that a bird
would stop only when it sees food and would not detour off course to save energy. It is
now known there are more factors at play than just the end destination (Alerstam 2001).
The other factors should be included in the creation of migration models, as I have done
61
here. The least accumulated-cost analysis does this by developing a cost structure to the
landscape. The end destination is still the primary motive as the analyses aim to find the
easiest route instead of the shortest.
Bird migration is one of the great spectacles of nature because an entire species
performs a mass exodus to a more favorable climate every year. It is unreasonable to
think an entire species will follow the same route during migration. Migration requires
stopovers to replenish energy by feeding. A single path across the continent could not
supply an entire species with enough food to complete the journey. Therefore multiple
paths are taken, creating a wide migration corridor. While the existing models show a
corridor it is not wide enough to include the range of migration paths used by Kirtland’s
Warblers. This least cost analysis created a model that does depict a much wider corridor
that allows for the variation of paths between individual birds. Kirtland’s Warblers are
particular about habitat and it can be assumed that a wide corridor is necessary to allow
for each bird to find its preferred habitat.
5.2 Golden-cheeked Warbler
Few records of migrating Golden-cheeked Warblers have been published. The
breeding and wintering ranges are known but only assumptions exist about the behavior
during migration. Most of the breeding range is in the hill country of Texas and the
wintering range is in the mountains with elevations well over 1000 meters. The species
seems to prefer hilly and mountainous terrain. Could we assume they would prefer it
during migration as well? A straight line between the two ranges extends directly across
the Gulf of Mexico. The results of the analysis show two possibilities. Following along
the east side of the mountains and crossing the gulf.
62
The strong association with mountainous terrain and the Golden-cheeked Warbler
imply that additional information should be included in analysis. Sightings of the
warblers have been recorded during migration along the Sierra Madre Oriental at
elevations (Groce et al. 2010). This would lead some to suspect the birds follow the
mountain range at heights ranging from 1,100 to 1,500 meters. The birds favor pine oak
forests which are common along the mountains at this elevation. An inclusion of the
preferred food sources of the Golden-cheeked Warbler could create a model that depicts
the use of the mountain range as its primary migration route. The analysis of the Golden-
cheeked Warbler is an example as to why every model must be adapted to a specific
species. The parameters set from the analysis of the Red-eyed Vireo were able to be
transferred to the Kirtland’s Warbler with only directional adjustments to the wind data
because of their similar nature.
The south east winds coming off the Gulf of Mexico may offer some insight into
the migration of the Golden-cheeked Warbler. With winds beating up against the
mountain range it is possible for large birds, such as Turkey Vultures, to use the draft to
gain elevation along the mountains (Mandel et al. 2011). It would then use gravity to
glide down the mountain side in its preferred direction. This would allow the bird to save
energy and also provides easy access to its more favorable food sources. While this is
possible for large winged birds, passerines are not usually seen gliding because of their
small wings.
The inclusion of the southeasterly winds in the Gulf of Mexico in the analysis
result in a model that supports the idea that Golden-cheeked Warblers may cross the gulf
during spring migration. The warblers are small birds and a trans-gulf crossing is difficult
63
but birds of similar size perform longer passages across the gulf with the aid of the same
winds. The only evidence that supports this theory are some records in the eBird database
of sightings near Corpus Christi, Texas during the spring migration months. Corpus
Christi is located where the birds would finish their passage. It is also a popular location
for bird watchers looking for migrating birds. This could explain the high number of
sightings but also the bird watchers are there because the migrants fly through Corpus
Christi.
5.3 Future Research
This analysis provides a proof of concept in that least accumulative cost analyses
can be used to model the migrations of birds. The success in creating maps that
reasonably depict known migrations of Red-eyed Vireo lead to building maps of
Kirtland’s Warbler migrations. While less information is known on the migration of the
Kirtland’s Warbler, the data currently available does show that the maps created in this
analysis could provide a new look on their migration patterns. The results of this study
are only a first step. More data and sophisticated parameterization would need to be
included in future analyses.
The slope raster was used as a basis for the rest of the resistance values. The slope
percentage may not have been the best way to quantify the barrier. The slope length
would have more accurately described the resistance to travel because it allows for the
calculation of the difference in length between flat ground and the sloped terrain. It may
also be beneficial to perform the analysis with a larger cell size. This would help to more
easily calculate the difference in elevation between cells. This would more generalize the
slope and eliminate the effect small scale elevation changes have on the slope values.
64
The wind data used in this study was based on historical averages from 1930 to
1996. The method for creating the wind data shapefiles allowed for error and could be
improved upon if not replaced. More recent and precise data should be utilized in future
research. Also, it is reasonable to think that wind pattern forecasts could be used in the
creation of a model that could predict future migration patterns.
It was noted earlier that fuel is one of the three most important factors in
determining flight patterns of birds (Alerstam 2011). This factor was not included in this
analysis. It may prove difficult to include this information. Different species all have their
own preferred habitats and their preferred nesting habitats may differ from stopover
points. Research must be done for each species so as to determine the preferred stopover
habitat and therefore assign the correct resistance value to land cover. Land cover can
also be very difficult to identify properly. There can be multiple layers of various land
covers at any given location. If the analysis was studying a raptor species then the ground
cover might be most important as it would lend more information as to what kind of
small game is inhabiting the area. Some land cover data would not be able to provide this
information so it is important to include data that depicts the land cover classifications
needed. There is a vast array of different land cover types and it would require
considerable time to go through an entire continents worth of classifications and assign
proper resistance values.
Few animals are able to supply themselves with energy by consuming the same
foods year round. Food sources appear at various times of the year and animals change
their diets accordingly. It is intuitive to place different values on wind direction based on
the time of year and direction of migration. The time of year also impacts the value of
65
various land cover types because birds may not get as much nutrition from those areas in
fall as they do in spring. An example would be how on a fall migration it would be
common to find migrating waterfowl to be feeding in cut corn fields during their fall
migration. The left over corn the birds are eating will not be present in those fields during
the return migration in the spring.
One reason for determining bird migration flyways is to prevent collisions with
tall building and towers. Some species have wide breeding and wintering ranges. The
analysis of the Red-eyed Vireo only studied the flyways from one section of the breeding
range to another section of the wintering range. In order to isolate the main
concentrations of migrating birds it would be necessary to include all of the breeding and
wintering ranges into the analysis. This could be done by performing the analysis as
before except multiple times. Once for each section of the breeding range to link to each
section of the wintering range. This is similar to the study of Black bear migration
performed by Cushman et al. (2009). They created least cost paths between 160 source
and 160 destination locations and averaged the results to create a migration corridor.
Instead of averaging paths to create a corridor, it would be averaging corridors to define a
more concentrated corridor.
The use of least accumulated-cost analyses for modeling bird migration patterns
can provide valuable information to preserving the natural environment. The
identification of high concentrations of migrating birds can assist in developing policies
that control where and how tall structures are built. The limitation of certain structure
heights, materials and designs can significantly reduce the number of migrating birds are
killed every year. This same information about flyway concentrations can be used to
66
identify areas of high concern for the preservation of critical stopover habitat. The least
accumulated-cost analysis could be performed at more local levels inside the already
identified corridors to determine exact land areas that need to be conserved. Human
development is constantly altering the environment and these changes can impact flight
patterns. When we destroy habitat birds will have to detour in order to find refueling sites
during their migration. The least accumulated-cost analysis could be done in advance of
development to determine the impact the project would have on the natural environment
and wildlife migration.

67
References
Able, Kenneth P. 1972. Fall migration in coastal Louisiana and the evolution of migration
patterns in the gulf region. The Wilson Bulletin 84 (3): 231–42.
Alerstam, Thomas. 2011. Optimal bird migration revisited. Journal of Ornithology 152
(1): 5.
Beier, Paul, Daniel R. Majka, and Shawn L. Newell. 2009. Uncertainty analysis of least-
cost modeling for designing wildlife linkages. Ecological Applications 19 (8):
2067–77.
Bocetti, Carol I., Deahn M. Donner and Harold F. Mayfield. 2014. Kirtland's Warbler
(Setophaga kirtlandii), The Birds of North America Online (A. Poole, Ed.).
Ithaca: Cornell Lab of Ornithology; Retrieved from the Birds of North America
Online: http://bna.birds.cornell.edu/bna/species/019 doi:10.2173/bna.19.
Chardon, J. P., Frank Adriaensen, and Erik Matthysen. 2003. Incorporating landscape
elements into a connectivity measure: A case study for the speckled wood
butterfly (Pararge aegeria L.). Landscape Ecology 18 (6): 561–73.
Chetkiewicz, Cheryl-Lesley B., and Mark S. Boyce. 2009. Use of resource selection
functions to identify conservation corridors. Journal of Applied Ecology 46 (5):
1036–47.
Cimprich, David A., Frank R. Moore and Michael P. Guilfoyle. 2000. Red-eyed Vireo
(Vireo olivaceus), The Birds of North America Online (A. Poole, Ed.). Ithaca:
Cornell Lab of Ornithology; Retrieved from the Birds of North America
Online: http://bna.birds.cornell.edu/bna/species/527 doi:10.2173/bna.527.
Clench, Mary Heimerdinger. 1973. The fall migration route of Kirtland's Warbler. The
Wilson Bulletin 85 (4) (12/01): 417–28.
Crawford, Robert L. 1980. Wind direction and the species composition of autumn TV
tower kills in northwest Florida. The Auk 97 (4) (10/01): 892–5.
Cushman, Samuel A., Kevin S. McKelvey, and Michael K. Schwartz. 2009. Use of
empirically derived source-destination models to map regional conservation
corridors. Conservation Biology 23 (2): 368–76.
Downs, J. A., and M. W. Horner. 2008. Spatially modelling pathways of migratory birds
for nature reserve site selection. International Journal of Geographical
Information Science 22 (6) (06/01; 2015/02): 687–702.
68
Driezen, Kassandra, Frank Adriaensen, Carlo Rondinini, C. Patrick Doncaster, and Erik
Matthysen. 2007. Evaluating least-cost model predictions with empirical dispersal
data: A case-study using radio tracking data of hedgehogs (Erinaceus
europaeus). Ecological Modelling 209 (2–4) (12/16): 314–22.
Elkins, Norman. 2004. Migrational drift and displacement. In Weather and bird
behavior. 3rd ed., 141A & C Black.
Epps, Clinton W., John D. Wehausen, Vernon C. Bleich, Steven G. Torres, and Justin S.
Brashares. 2007. Optimizing dispersal and corridor models using landscape
genetics. Journal of Applied Ecology 44 (4): 714–24.
Etherington, Thomas R, and E. Penelope Holland. 2013. Least-cost path length versus
accumulated-cost as connectivity measures. Landscape Ecology 28 (7) (08/01):
1223–9, http://dx.doi.org/10.1007/s10980-013-9880-2.
Fiedler, Wolfgang. 2009. New technologies for monitoring bird migration and
behaviour. Ringing & Migration 24 (3) (01/01; 2015/08): 175–
9. http://dx.doi.org/10.1080/03078698.2009.9674389.
Gehring, Joelle, Paul Kerlinger, and Albert M. Manville. 2011. The role of tower height
and guy wires on avian collisions with communication towers. The Journal of
Wildlife Management 75 (4): 848–55.
———. 2009. Communication towers, lights, and birds: Successful methods of reducing
the frequency of avian collisions. Ecological Applications 19 (2) (03/01;
2015/02): 505–14.
Goode, J. P. 1925. The homolosine projection: A new device for portraying the earth's
surface entire. Annals of the Association of American Geographers 15 (3) (09/01;
2015/02): 119–25.
Groce, J. E., H. A. Mathewson, M. L. Morrison, and R. N. Wilkins. 2010. Scientific
evaluation for the 5-year status review of the Golden-cheeked Warbler. Final
report submitted to the Texas Parks and Wildlife Department, Austin, Texas, and
to the U.S. Fish and Wildlife Service, Region 2, Albuquerque, New Mexico,
USA.
Gudmundsson, Gudmundur A., and Thomas Alerstam. 1998. Optimal map projections
for analyzing long-distance migration routes. Journal of Avian Biology 29 (4)
(12/01): 597–605.
Jr., Daniel Klem. 2009. Preventing bird-window collisions. The Wilson Journal of
Ornithology 121 (2) (06/01): 314–21.
69
Kautz, Randy, Robert Kawula, Thomas Hoctor, Jane Comiskey, Deborah Jansen, Dawn
Jennings, John Kasbohm, et al. 2006. How much is enough? Landscape-scale
conservation for the Florida Panther. Biological Conservation 130 (1) (6): 118–
33.
Klem, Daniel, Jr. 1990. Collisions between birds and windows: Mortality and
prevention. Journal of Field Ornithology 61 (1) (01/01): 120–8.
Klem, Daniel, Christopher J. Farmer, Nicole Delacretaz, Yigal Gelb, and Peter G.
Saenger. 2009. Architectural and landscape risk factors associated with bird-glass
collisions in an urban environment. The Wilson Journal of Ornithology 121 (1)
(03/01; 2015/02): 126–34.
Kuenzi, Amy Jo, Frank R. Moore, and Ted R. Simons. 1991. Stopover of neotropical
landbird migrants on east Ship Island following trans-gulf migration. The
Condor 93 (4) (11/01): 869–8.
Li, Hailong, Dihua Li, Ting Li, Qing Qiao, Jian Yang, and Hemin Zhang. 2010.
Application of least-cost path model to identify a Giant Panda dispersal corridor
network after the Wenchuan earthquake—Case study of Wolong nature reserve in
china. Ecological Modelling 221 (6) (3/24): 944–52.
Longcore, Travis, Catherine Rich, and Sidney A. Gauthreaux Jr. 2008. Height, guy wires,
and steady-burning lights increase hazard of communication towers to nocturnal
migrants: A review and meta-analysis. The Auk 125 (2) (04/01): 485–92.
Longcore, Travis, Catherine Rich, Pierre Mineau, Beau MacDonald, Daniel G. Bert,
Lauren M. Sullivan, Erin Mutrie, et al. 2013. Avian mortality at communication
towers in the United States and Canada: Which species, how many, and
where? Biological Conservation 158 (0) (/2): 410–19.
Longcore, Travis, Catherine Rich, Pierre Mineau, Beau MacDonald, Daniel G. Bert,
Lauren M. Sullivan, Erin Mutrie, et al. 2012. An estimate of avian mortality at
communication towers in the United States and Canada. Plos One 7 (4) (02/20):
e34025.
Loss, Scott R., Tom Will, Sara S. Loss, and Peter P. Marra. 2014. Bird-building
collisions in the United States: Estimates of annual mortality and species
vulnerability. The Condor 116 (1) (02/01; 2015/02): 8–23.
Mandel, James T., Gil Bohrer, David W. Winkler, David R. Barber, C. S. Houston, and
Keith L. Bildstein. 2011. Migration path annotation: Cross-continental study of
migration-flight response to environmental conditions. Ecological Applications 21
(6) (09/01; 2015/02): 2258–68.
70
Mayfield, H.F. 1972. Winter habitat of Kirtland’s Warbler. Wilson Bulletin 84(3):347–
349.
Meegan, Rebecca P., and David S. Maehr. 2002. Landscape conservation and regional
planning for the Florida Panther. Southeastern Naturalist 1 (3) (09/01; 2015/02):
217–32.
Meyer, Susanna K., Reto Spaar, and Bruno Bruderer. 2000. To cross the sea or to follow
the coast? Flight directions and behaviour of migrating raptors approaching the
Mediterranean Sea in autumn. Behaviour 137 (3) (03/01): 379–99.
Moore, Frank R., Paul Kerlinger, and Ted R. Simons. 1990. Stopover on a gulf coast
barrier island by spring trans-gulf migrants. The Wilson Bulletin 102 (3) (09/01):
487–500.
Nakamura, Hiroshi, Yoshitomo Miyazawa, and Kenichi Kashiwagi. 2005. Behavior of
radio-tracked Common Cuckoo females during the breeding season in
Japan. Ornithological Science 4 (1) (03/01; 2015/08): 31–
41, http://dx.doi.org/10.2326/osj.4.31.
Purcell, Jessica, and Anders Brodin. 2007. Factors influencing route choice by avian
migrants: A dynamic programming model of Pacific Brant migration. Journal of
Theoretical Biology 249 (4) (12/21): 804–16.
Rabinowitz, Alan, and Kathy A. Zeller. 2010. A range-wide model of landscape
connectivity and conservation for the Jaguar, (Panthera onca). Biological
Conservation 143 (4) (4): 939–45.
Sandberg, Roland, and Frank R. Moore. 1996. Migratory orientation of Red-eyed Vireos,
(Vireo olivaceus), in relation to energetic condition and ecological
context. Behavioral Ecology and Sociobiology 39 (1) (07/01): 1–10.
Sawyer, Sarah C., Clinton W. Epps, and Justin S. Brashares. 2011. Placing linkages
among fragmented habitats: Do least-cost models reflect how animals use
landscapes? Journal of Applied Ecology 48 (3): 668–78.
Sillett, T. Scott, and Richard T. Holmes. 2002. Variation in survivorship of a migratory
song-bird throughout its annual cycle. Journal of Animal Ecology 71 (2): 296–
308.
Steinwand, D. R. 1994. Mapping raster imagery to the Interrupted Goode Homolosine
projection. International Journal of Remote Sensing 15 (17) (11/01; 2015/02):
3463–71.
Steinwand, Daniel R., John A. Hutchinson, and J. P. Snyder. 1995. Map projections for
global and continental data sets and an analysis of pixel distortion caused by
71
reprojection. Photogrammetric Engineering and Remote Sensing 61 (12): 1487–
97, http://www.asprs.org/Photogrammetric-Engineering-and-Remote-Sensing/PE-
RS-Journals.html.
U. S. Fish and Wildlife Service. 2014. Multi Species Recovery Plan.
http://www.fws.gov/verobeach/MSRPPDFs/Kirtlandswarbler.pdf.
Usery, Lynn E., and Jeong Chang Seong. 2001. All equal-area map projections are
created equal, but some are more equal than others. Cartography and Geographic
Information Science 28 (3) (01/01; 2015/02): 183–94.
Walker, Richard, and Lance Craighead. 1997. Analyzing wildlife movement corridors in
Montana using GIS. Paper presented at ESRI User Conference.
Wand, Ian J., Wesley K. Savage, and H. Bradley Shaffer. 2009. Landscape genetics and
least-cost path analysis reveal unexpected dispersal routes in the California Tiger
Salamander (Ambystoma californiense). Molecular Ecology 18 (7): 1365–74.

Abstract (if available)

Abstract Millions of birds are killed every year during their annual migration by colliding with tall communication towers and buildings. The goal of this study is to identify areas of specific concern for avian species during migration by modeling potential migration corridors for Red-eyed Vireo (Vireo olivaceus), Kirtland’s Warbler (Setophaga kirtlandii), and Golden-cheeked Warbler (Setophaga chrysoparia) as a case study. These avian species perform transcontinental migrations each year. This study uses a least accumulated-cost analysis to predict probability of use of routes between winter and summer ranges by analyzing the presumed energetic cost of changing altitude (in response to topographic relief), traversing large bodies of water, and compensating for wind. Previous descriptions of migration pathways depict straight lines that do not take into account geographic barriers. This study compares the results of existing methods to the least accumulative cost model. The completion of the analysis on Red-eyed Vireo allows the same analysis to be performed on two more rare species, the Kirtland’s Warbler and the Golden-cheeked Warbler. The results of this study show that least accumulated cost analyses are a viable option to assisting in determining preferred migration routes for migratory birds. Least accumulated-cost analyses demand significant computing resources, which can prevent studies of this size from being performed. Advances in technology now enable studies of this magnitude to be performed and this study is a proof-of-concept to illustrate the potential benefits of integrating these analyses into conservation planning.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Modeling geopolitics in Tikal through least cost paths

PDF

Applying least cost path analysis to search and rescue data: a case study in Yosemite National Park

PDF

Archaeological least cost path modeling: a behavioral study of Middle Bronze Age merchant travel routes across the Amanus Mountains, Turkey

PDF

Assessing the connectivity for the jaguar (Panthera onca) in the United States-Mexico border ecoregions using species distribution modeling and factorial least cost path analysis

Asset Metadata

Creator Winters, William (author)

Core Title Identifying areas of high risk for avian mortality by performing a least accumulated-cost analysis

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 09/17/2015

Defense Date 09/01/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag avian mortality,Birds,corridor,cost distance,DEM,GIS,Golden-cheeked Warbler,Interrupted Goode Homolosine projection,Kirtland's Warbler,least accumulated-cost,least-cost,migration,OAI-PMH Harvest,Red-eyed Vireo,resistance values,slope,Water,wind

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Longcore, Travis (committee chair), Kemp, Karen (committee member), Wilson, John P. (committee member)

Creator Email wjwinters4@gmail.com,wwinters@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-184901

Unique identifier UC11273063

Identifier etd-WintersWil-3930.pdf (filename),usctheses-c40-184901 (legacy record id)

Legacy Identifier etd-WintersWil-3930.pdf

Dmrecord 184901

Document Type Thesis

Format application/pdf (imt)

Rights Winters, William

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA