Coyote (Canis Latrans) Occurrence Relative to Human Use on Glenbow Ranch Provincial Park, Alberta

University of Calgary PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2016 Coyote (Canis latrans) occurrence relative to human use on Glenbow Ranch Provincial Park, Alberta

Lantz, Jamie

Lantz, J. (2016). Coyote (Canis latrans) occurrence relative to human use on Glenbow Ranch Provincial Park, Alberta (Unpublished master's thesis). University of Calgary, Calgary, AB. doi:10.11575/PRISM/25479 http://hdl.handle.net/11023/2864 master thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca UNIVERSITY OF CALGARY

Coyote (Canis latrans) occurrence relative to human use on

Glenbow Ranch Provincial Park, Alberta

Jamie Lantz

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF MASTER OF SCIENCE

GRADUATE PROGRAM IN GEOGRAPHY

CALGARY, ALBERTA

March, 2016

© Jamie Lantz 2016 Abstract Although coyote (Canis latrans) attacks in Canada are rare, they tend to fuel lethal action. However, killing coyotes is ineffective for reducing long-term conflict and is ecologically destructive. Thus, coexistence is crucial. Understanding how humans and coyotes share the landscape is important to help mitigate conflict. I explored human- coyote co-occurrence in Glenbow Ranch Provincial Park, Alberta between June 2014 and June 2015. Cameras and scat surveys collected data on high and low human use trails. These two methods detected coyote occurrence equally on both trail types as well as across seasons. On all trail types coyote occurrence was greatest during the winter (versus fall and summer), dispersal season (versus the weaning season) and nighttime and daytime (versus the twilight periods). Coyote occurrence significantly decreased on trails when cyclists, vehicles and prey were present. In contrast, coyote occurrence increased when coyote(s) used trails within the past day.

ii Acknowledgements I would like to thank my supervisor, Dr. Shelley Alexander, for her guidance and knowledge throughout my degree. Additionally, I would like to thank my committee members, Dr. Dianne Draper and Dr. Mary Pavelka. As well, thank you to Dr. Darren Bender for sitting on my proposal defence committee. Thank you to my lab mates, Victoria Lukasik, Elisabeth van Stam, Kyle Plotsky and Jay Reid, for their support. Thanks to Derek Wilson for his help in the lab as well as Bart Hulshof for his knowledge of GIS. Thank you to Dr. Tak Fung for helping me during the statistical analysis. Also, thank you to the Geography Department students, teachers and administrative staff for their encouragement. Thank you to those who provided me with the resources necessary to complete my degree. Thank you to the Natural Sciences and Engineering Research Council of Canada (NSERC), the ACA Grants in Biodiversity (supported by the Alberta Conservation Association) and the University of Calgary for funding. Thank you to the Miistakis Institute for lending me remotely-triggered cameras. Thanks to Alberta Parks for providing me with important camera data, map layers and access to Glenbow Ranch Provincial Park. As well, thank you to the Glenbow Ranch Provincial Park Foundation for their wealth of information. Thank you to the Conservation Officers of the park for helping me set up and take down my cameras, including Huntley Johnston, Curtis Haslehurst and John Duguid. Special thanks to my family and friends. I would not have been able to do this without you.

iii Dedication To my family and friends: To the field we went, my family and friends In search of coyote feces. Jarrett, Lia, Vic, Anne, Paul and Jen Helped support me during my thesis.

Jarrett focused on bike “wheelies” And dropped batteries into the snow. Lia brought Brody, her collie-cross Who trail walked like a torpedo.

Anne and Paul braved wildlife paths, Jen and I never found poop. Victoria came with me at the start We got lost north of Yodel Loop.

Uncle Mark joined in as well as my dog They really gave it their all. Mom came with me countless times Walking the low use trails long haul.

Dad came with us on the worst day The snow was up to our hips. Although I broke trail dad still complained His heart rhythm was on the blitz.

Thank you everyone for being there Defying snow, cold, rain and wind. Maybe we’ll go back in a little while, but for now… An adventure it’s truly been.

iv Table of Contents Abstract ...... ii Acknowledgements ...... iii Dedication ...... iv Table of Contents ...... v List of Tables ...... vii List of Figures and Illustrations ...... viii List of Symbols, Abbreviations and Nomenclature ...... ix Epigraph ...... xi

Chapter One: Introduction ...... 1 1.1 Project Summary ...... 1 1.2 Problem Statement and Research Questions ...... 2 1.3 Study Area ...... 3 1.4 Thesis Organization ...... 6

CHAPTER TWO: BACKGROUND ...... 7 2.1 Coyote Natural History ...... 7 2.1.1 Taxonomy and Appearance ...... 7 2.1.2 Distribution and Habitat ...... 8 2.1.3 Territory and Home Range Size ...... 9 2.1.4 Diet ...... 10 2.1.5 Reproduction ...... 11 2.1.6 Social Structure ...... 12 2.1.7 Competitors ...... 13 2.2 Human-Coyote Conflict ...... 14 2.3 Ecological Importance of Coyotes ...... 18 2.4 Coexistence ...... 18

CHAPTER THREE: COMPARING COYOTE OCCURRENCE BETWEEN CAMERA TRAPPING AND SCAT SURVEYS ...... 21 3.1 Introduction ...... 21 3.1.1 Literature Review ...... 21 3.1.2 Significance ...... 23 3.2 Research Question and Hypotheses ...... 24 3.3 Study Area ...... 24 3.4 Methods ...... 26 3.4.1 Statistical Justification and Analysis ...... 29 3.5 Results ...... 30 3.6 Discussion ...... 31 3.7 Conclusion ...... 34

CHAPTER FOUR: COYOTE OCCURRENCE RELATIVE TO PEOPLE IN GRPP .....35 4.1 Introduction ...... 35 4.1.1 Literature Review ...... 35 4.1.2 Significance ...... 35

v 4.2 Research Question, Objectives and Hypotheses ...... 36 4.3 Study Area ...... 37 4.4 Methods ...... 37 4.4.1 Statistical Justification and Analysis ...... 41 4.4.1.1 Research Objective 1 ...... 41 4.4.1.2 Research Objective 2 ...... 43 4.5 Results ...... 47 4.5.1 Research Objective 1 ...... 47 4.5.2 Research Objective 2 ...... 52 4.6 Discussion ...... 57 4.6.1 Research Objective 1 ...... 57 4.6.2 Research Objective 2 ...... 59 4.7 Conclusion ...... 63

CHAPTER FIVE: MANAGEMENT RECOMMENDATIONS AND FUTURE RESEARCH ...... 65 5.1 Management Recommendations ...... 65 5.2 Future Research ...... 66

REFERENCES ...... 68

APPENDICES ...... 78

vi List of Tables Table 3.1 Number of two-week periods between September 11, 2014 and November 26, 2014 that detected coyote occurrence based on method and trail type...... 30

Table 3.2 Number of two-week periods between September 11, 2014 and June 16, 2015 that detected coyote occurrence using various method types across different seasons...... 31

Table 4.1 Variables used in the binary multiple logistic regression...... 39

Table 4.2 Skewness and kurtosis values of the data...... 43

Table 4.3 Models created for binary multiple logistic regression...... 46

Table 4.4 Comparison of coyote occurrence between trail types within each season...... 48

Table 4.5 Comparison of coyote occurrence between trail types within the weaning and dispersal life cycle stages...... 49

Table 4.6 Comparison of coyote occurrence between trail types within time periods of the day...... 51

Table 4.7 Average number of coyotes/week on high and low human use trails during different timeframes...... 52

Table 4.8 Comparison of models 1 to 14...... 53

vii List of Figures and Illustrations Figure 1.1 Glenbow Ranch Provincial Park...... 4

Figure 1.2 Glenbow Ranch Provincial Park...... 5

Figure 2.1 Coyote...... 8

Figure 3.1 High and low human use trail systems in GRPP...... 25

Figure 4.1 Winsorization process...... 42

Figure 4.2 Comparison of harmonized mean coyote occurrence per week between summer, fall and winter on all trail types...... 47

Figure 4.3 Comparison of harmonized mean coyote occurrence per week between the weaning period and the dispersal season on all trail types...... 49

Figure 4.4 Comparison of harmonized mean coyote occurrence per week between morning twilight, daytime, evening twilight and nighttime periods on all trail types...... 50

Figure 4.5 Comparison of harmonized mean coyote occurrence per week between high and low human use trails during the nighttime...... 51

viii List of Symbols, Abbreviations and Nomenclature

Symbol Definition

GRPP Glenbow Ranch Provincial Park

BMLR Binary multiple logistic regression

Low human use trails Consist of wildlife trails park visitors are not permitted to use. Although visitors must remain on designated paths (high human use trails), not all people comply.

High human use trails Consist of paved and gravel trails park visitors are allowed to use. Cycling is permitted only on designated high human use trails.

Natural habitat/areas/spaces Are “large, contiguous areas of natural vegetation” (Riley et al., 2003, p. 569) used for wildlife preservation. Some of these areas may also permit low to moderate levels of human recreation, such as hiking and camping. Examples include provincial parks, national parks and preserves.

Modified habitat/areas/spaces Have moderate levels of human use. Modified habitat is generally smaller than natural habitat and does not contain 100% natural vegetation. Examples include golf courses and gardens.

ix Urban areas Have high levels of human use. Urban areas may contain natural spaces and modified habitat. Examples include cities and towns.

Human-coyote conflict Encompasses negative interactions involving aggressive coyote behavior, such as growling, as well as physical contact, such as biting (Lukasik & Alexander, 2011).

x Epigraph

Unless someone like you cares a whole awful lot, nothing is going to get better. It’s not.

- Dr. Seuss, The Lorax

Chapter One: Introduction

1.1 Project Summary

I examined human and coyote (Canis latrans) occurrence in Glenbow Ranch Provincial Park (GRPP), Alberta. This research is important because the park is located between two growing urban centers, Calgary and Cochrane. As a result, human pressure surrounding GRPP is expected to increase. Obtaining baseline data on human-coyote co- occurrence may help with management and coexistence. To date, there has been one report of visitor-coyote conflict in GRPP (Johnston, 2015). This occurred in March 2012. A visitor walking their dog encountered a coyote that was “blocking the path and seemed to be focused on the dog”. Only after “a good deal of yelling” did the coyote move off the trail. The coyote’s teeth were not bared, “no contact was made” and the occurrence took place near “active den sites”. During the same year a rancher witnessed coyotes attacking and killing one calf in the park. After the attack, coyotes “attempted” approaching cows and calves, however no further attacks were made. Increased conflict between people and coyotes may be related to human population growth (Timm, Baker, Bennett, & Coolahan, 2004). In 2013 the Canadian populace was 35.2 million (Statistics Canada, 2014b). By 2063 it is expected to grow by approximately 15.8 million. Out of the Canadian provinces, Alberta is predicted to grow the most with a population increase of roughly 2.2 million by 2038. Alberta’s fastest growing Census Metropolitan Area is Calgary, which grew by 3.6% between 2013 and 2014 (Statistics Canada, 2015) and reached a population of 1 195 194 (The City of Calgary, 2014). To the west of Calgary, Cochrane is also expanding (Town of Cochrane, 2015). In 2000 Cochrane’s population was 11 173, which grew to 23 084 in 2015. Human population growth is problematic for wildlife because it leads to a decrease in natural habitat. In 2006 Calgary occupied 726.5 km2 (Statistics Canada, 2010). Over the course of 5 years this area grew to 825.29 km2 (Statistics Canada, 2014a). Consequently, 98.79km2 of habitat was removed and incorporated into the city; wildlife

would have had to re-locate or adjust to an urban environment. Such habitat loss is the leading threat to species in Canada (Venter et al., 2006). Some species are able to adapt to urbanization, such as the coyote. For example, it is estimated that approximately 2 000 coyotes live in the Chicago metropolitan area (Dell'Amore, 2014). Urban areas provide coyotes with water, such as ponds, as well as habitat found in parks and treed spaces (Morey, Gese, & Gehrt, 2007). Cities also provide coyotes with food. They predate small rodents, rabbits, insects and birds and will also consume pet food, domestic animals, fruits and garbage (Lukasik & Alexander, 2012; Morey et al., 2007). Unfortunately, living in close proximity to people has led to human-coyote conflict (Alexander & Quinn, 2011). Human-coyote conflict encompasses negative interactions involving aggressive coyote behavior, such as growling, as well as physical contact, such as biting (Lukasik & Alexander, 2011). Some people believe that untargeted lethal control can reduce or avert these negative interactions (CBC News, 2014). However, killing coyotes is detrimental because coyotes are ecologically important (Crabtree & Sheldon, 1999b; Crooks & Soulé, 1999; Henke & Bryant, 1999; Young & Jackson, 1951) and lethal control is ineffective in the long term (Fox & Papouchis, 2005). Thus, it is crucial to understand how coyotes and humans can coexist to better manage conflict while accommodating coyote survival needs.

1.2 Problem Statement and Research Questions

This study used non-invasive methods to explore how people and coyotes co- occurred in GRPP. It also provided baseline data should urban expansion intensify around the park. This research project was driven by a joint interest of the Foothills Coyote Initiative, Alberta Parks and the Glenbow Ranch Provincial Park Foundation to better understand human-coyote interactions and coyote diet in GRPP.

General Research Questions:

Research Question 1: Do remotely triggered cameras, scat surveys and a combination of these two methods detect coyote occurrence equally on high and low human use trails as well as across different seasons?

Research Question 2: How do coyotes and humans co-occur spatio-temporally within GRPP on high and low human use trails? This question was broken into two objectives:

Research Objective 1: Is coyote occurrence less on high human use trails compared to low human use trails? Does coyote trail occurrence fluctuate across different seasons, life cycle stages and times of the day?

Research Objective 2: How do human-related disturbances, past coyote trail occurrence, prey, moon phase, season, life cycle stage and time of day relate to coyote occurrence on trails?

1.3 Study Area

Glenbow Ranch (51°10’6.96’’N, -114°23’34.44’’W) became a designated provincial park in 2008 (Alberta Parks, 2015). GRPP occupies an area of 1 314 hectares and is located between Calgary and Cochrane, Alberta (Figure 1.1) (Stenson, 2012). The Bow River traverses through the south of GRPP. Highway 1A is situated north of the park connecting Cochrane to Calgary. GRPP is characteristic of the foothills region with rolling hills and valleys. The park contains both grassland and parkland habitats (Figure 1.2). Some of the common plants include: wolf willow (Elaeagnus commutata), Canada buffaloberry (Shepherdia canadensis), prairie crocus (Anemone patens), ground juniper (Juniperus communis), aspen (Populus tremuloides) and fescue grasses (Glenbow Ranch Park Foundation, n.d.-a). Various animals inhabit GRPP, such as mule deer (Odocoileus

hemionus), moose (Alces alces), North American porcupines (Erethizon dorsatum), American badgers (Taxidea taxus), Richardson’s ground squirrels (Urocitellus richardsonii), American crows (Corvus brachyrhynchos), red foxes (Vulpes vulpes) and coyotes (Canis latrans) (Glenbow Ranch Park Foundation, n.d.-b). Cattle roam in designated areas of the park as GRPP contains active cattle ranches (Stenson, 2012).

Figure 1.1 Glenbow Ranch Provincial Park.

Figure 1.2 Glenbow Ranch Provincial Park. From GeoKs (2012).

Summer months in GRPP have an average temperature of 15.3°C with approximately 72.2 mm of precipitation (Environment Canada, 2015). In fall the average temperature is 4.6°C with 24.5 mm of precipitation. Winter and spring have average temperatures of -6.4°C and 4.2°C as well as precipitation levels of 9.7 mm and 33.3 mm respectively. The region also experiences Chinooks, which bring dry, warm air to the park (Environment Canada, 2013). GRPP opened for public use in August 2011 (Stenson, 2012). The park consists of over 25 km of paved and gravel paths (Alberta Parks, 2015). Activities include: hiking, cycling and bird watching. People may walk dogs in the park. By law, dogs must be kept on leash, however not all dog-walkers comply.

1.4 Thesis Organization

This is a manuscript-based thesis organized by chapter. Chapter 1 outlines the research questions and study area used in Chapters 3 and 4. Chapter 2 provides background literature on coyote natural history, human-coyote conflict and coyote ecological importance. The following chapter compares the relative ability of camera trapping, scat surveys and a combination of these two methods to detect coyote occurrence. Chapter 4 investigates coyote occurrence across seasons and between high and low human use trails. This chapter also analyzes how different factors, such as human-related disturbances and season, relate to coyote trail occurrence. Chapter 5 summarizes the key findings of Chapters 3 and 4. Management strategies are also suggested.

Chapter Two: Background

2.1 Coyote Natural History

Understanding coyote natural history is important to this study. Knowledge of coyotes can help establish methods. For example, home range size is often used to determine appropriate distances between cameras (Kelly & Holub, 2008; Silver, 2004). Natural history may also provide insight on experimental outcomes. For instance, coyotes can be territorial (Bowen, 1981; Camenzind, 1978; Fox & Papouchis, 2005). I would assume that this behavioural trait might increase pack member trail occurrence if “intruding” coyotes are detected. This study also investigates coyote occurrence relative to prey occurrence, which is why familiarity of prey species is important. Additionally, knowledge of coyote reproduction and social structure may help shed light on coyote occurrence during the different life cycle stages explored in this research. Thus, understanding coyote ecology, behavior and adaptations may help with experimental set- up and interpreting results.

2.1.1 Taxonomy and Appearance

Coyotes (Figure 2.1) belong to the family Canidae, which consists of 36 species worldwide (IUCN, 2015). Average length is 1.0 to 1.5 metres, with males larger than females (Bekoff & Gese, 2003). Weight is approximately 18-30 pounds (8-14 kg) (Young & Jackson, 1951). Most coyotes have a grey pelage with a reddish hue and pale underside (Bekoff & Gese, 2003). North-ranging coyotes tend to be darker with a longer, thicker coat compared to south-ranging coyotes, which have a more fulvous, lighter coat (Young & Jackson, 1951). They are similar in appearance to gray phase wolves, however coyotes are usually smaller (Fox & Papouchis, 2005). As well, coyotes have larger ears and a more pointed snout relative to their face.

Figure 2.1 Coyote. From (Martin, 2012).

2.1.2 Distribution and Habitat

Coyotes evolved on the North American continent and have been present in their current form (latrans) for 1 million years (Wang & Tedford, 2008). However, fossil evidence has been sporadic, making it difficult to accurately identify historic distribution (Alexander, 2014). Some literature suggests coyotes historically occupied the prairies in southern Canada, states west of the Mississippi river and northern Mexico (Young & Jackson, 1951). One hypothesis is that human modification of the landscape into pastureland for livestock allowed coyotes to increase their range (Moore & Parker, 1992). Not only were coyotes able to consume domesticated animals (Young & Jackson, 1951), but based on arguments of competitive exclusion, coyotes are believed to have moved into human-inhabited regions because these areas dispelled competitors such as cougars (Puma concolor) and wolves (Canis lupus) (Berger & Gese, 2007; Gompper, 2002; Moore & Parker, 1992). Thus, it is possible that the consequences of human development allowed coyote populations to expand.

Currently, coyotes inhabit the continental United States, Alaska, most of Canada, Mexico and parts of Central America (Moore & Parker, 1992). Coyotes occupy a variety of habitat types, such as the tropics, mountains, forests and even urban areas (Sillero- Zubiri, Hoffmann, & Macdonald, 2004). Within urban locations coyotes tend to select for natural spaces, which are defined as “large, contiguous areas of natural vegetation” (Riley et al., 2003, p. 569). For the purpose of this research, natural spaces include areas that have zero to moderate levels of human activity such as provincial/national/state parks as well as nature preserves. Natural spaces provide habitat for wildlife and may also offer recreational opportunities for visitors. GRPP is an example of a natural space; it is a provincial park that protects native wildlife, while providing the public with trail systems to hike, cycle and walk dogs.

2.1.3 Territory and Home Range Size

Territory is defined as “any defended area” (Noble, 1939, p. 267), while a home range is the “area traversed by the individual in its normal activities of food gathering, mating, and caring for young” (Burt, 1943, p. 351). Unlike lone coyotes, coyotes living in a pack or in pairs have a territory (Bowen, 1981; Camenzind, 1978; Fox & Papouchis, 2005). Territories are established through direct defense, howling and scent-marking (Gese, 2004). The latter involves urination, defecation and ground scratching. In general, coyotes scent-mark their territory at least once every hour. Although all pack members help to maintain the territory, the alpha pair plays the dominant role in territory protection and boundary marking. Home range and territory sizes vary depending on habitat type and level of human disturbance. In the Athabasca River Valley (ARV), a natural area north of Jasper, Alberta, coyote territory size estimates were between 8.0 km2 and 20.0 km2 (Bowen, 1978). Home range size was 13.7 km2 (Bowen, 1982). Coyotes living in farmland regions have similar home range sizes to coyotes residing in natural areas. In the Champlain Valley, Vermont coyote home range was approximately 17.9 km2 (Person & Hirth, 1991). The study area contained roughly 40% forest and 60% farmland. Another study site in northwestern Texas consisted of farmland, native prairie and conservation areas to

protect vegetation from agriculture (Kamler, Ballard, Lemons, Gilliland, & Mote, 2005). There, resident coyote home range was approximately 10.1 km2. Unlike natural areas and farmland regions, urban environments seem to have a larger variation in coyote home range and territory sizes. In an urban area west of Los Angeles, California adult coyote home range averaged 4.96 km2 (Riley et al., 2003). Likewise, resident coyote home range and territory sizes in the Chicago metropolitan area were 4.95 km2 (Gehrt, Anchor, & White, 2009; Urban Coyote Research, 2015). Within these two previous studies, most coyotes selected for natural spaces, avoiding contact with people. In contrast, in the Banff town site, Alberta, coyote home range was between 16.2 km2 and 39.8 km2 for resident coyotes (Gibeau, 1993). Furthermore, coyotes in Banff used spaces regardless of human occurrence, except in the summer. Throughout summer months coyotes avoided high human-use areas during the day. I was unable to find coyote home range and territory sizes in habitat types similar to GRPP. However, because GRPP contains natural areas, farmland and human recreation sites, coyote home range would likely be a rough average of the three habitat types. Since there is a high amount of human use in the park I estimate that coyote home range could be between 5 km2 and 8 km2.

2.1.4 Diet

Coyotes are opportunistic omnivores (Fox & Papouchis, 2005). They consume ungulates, hare and rodents (Crabtree & Sheldon, 1999b) as well as fruit and insects (Bekoff, 1977). Coyotes may also hunt ungulate species such as deer and elk (Gese & Grothe, 1995). Although they usually predate older, sicker individuals, they will also attack healthy members of the population (Crabtree & Sheldon, 1999b). During winter coyotes may feed heavily on carrion. In the summer and fall, rodents make up a large part of their diet. Coyotes also depredate livestock (Fall, 1990) and will consume livestock carcasses left by ranchers (Danner & Smith, 1980). Depredation of domesticated animals is most common when human disturbance has decreased natural prey abundance, which may cause coyotes to turn to livestock for sustenance (Fox & Papouchis, 2005). In urban

areas coyotes may feed on pets as well as refuse (Alexander & Quinn, 2011; Fox & Papouchis, 2005; Lukasik & Alexander, 2012). Coyote diet studies have been performed in and around Calgary. Small mammals comprised 76.0% of urban and rural coyote scats (Fortin-McCuaig, 2012). Plants (39.9%), large mammals (16.6%), medium-sized mammals (9.2%), birds (8.6%), crab apples (8.6%), insects (5.3%) and berries (1.8%), as well as human-related foods including garbage (8.0%), cattle (3.3%) and domestic animals (2.1%) were also consumed. An earlier study performed in Calgary demonstrated similar results (Lukasik & Alexander, 2012). Urban coyote diet consisted primarily of small mammals, but also contained plants, fruit and garbage. Six of the 484 scats contained domestic animal remains. In the Calgary area, rural coyotes consumed more small mammals than urban coyotes (Fortin-McCuaig, 2012). Additionally, urban coyotes did not consume cattle, whereas 6.2% of rural coyote scat contained cattle remains. Urban coyote diet also appeared more diversified than rural coyote diet. Coyotes living within Calgary consumed more birds, insects, plants, crab apples, garbage and pets than did rural-living coyotes. Coyote diet specific to GRPP is in the process of being analyzed using samples collected contemporaneously with this study (Alexander, 2015).

2.1.5 Reproduction

Females reach sexual maturity when they are 10 months old (Crabtree & Sheldon, 1999b). However, age of reproduction may vary depending on level of persecution and prey density. In areas where human-related coyote mortality is high (exploited areas), females may reproduce as yearlings (Fox & Papouchis, 2005). Conversely, in unexploited regions, it is more common for females to reproduce between ages 2 and 5 (Crabtree & Sheldon, 1999b). Although the pack’s hierarchal structure usually ensures only the alpha pair mate (Fox & Papouchis, 2005), more females may breed in exploited areas due to pack fragmentation (Fox & Papouchis, 2005). Likewise, more females may breed during times of high prey abundance (Gier, 1968).

Estrous takes place once a year in early February when mating occurs (Crabtree & Sheldon, 1999b). At the end of March, pregnant females begin digging 2 to 3 dens. Dens are usually located on banks/slopes with a southerly aspect or in thickets or logs (Gier, 1968). At the beginning of April, after approximately 63 days of gestation (Bekoff, 1977), pups are born (Crabtree & Sheldon, 1999b). Females give birth to roughly 6 pups, which are helpless and blind (Bekoff, 1977). Litter size may be smaller during times of low prey abundance and high coyote densities (Gier, 1968; Knowlton, 1972). For example, in South Texas areas of low coyote density had an average litter size of 6.9 pups (Knowlton, 1972). In contrast, high-density areas had an average litter size of 4.3 pups. Adults may move their young between dens if they detect human disturbance and if appropriate undisturbed areas are available (Harrison & Gilbert, 1985). When coyotes are approximately 2-3 weeks old, they start to venture out of the den (Bekoff, 1977). By mid- May, roughly 5-6 weeks old, pups are weaned and food is obtained from the adults by regurgitation (Crabtree & Sheldon, 1999b). In central Alberta it was estimated that roughly 71% of coyotes living in the wild die between ages 0 and 1 (Nellis & Keith, 1976). Mortality rates decrease after one year of age to approximately 36-42%. The oldest coyotes discovered by Nellis and Keith (1976) in central Alberta were a 13 ½ year-old male and an 11 ½ year-old female.

2.1.6 Social Structure

Coyotes may live in groups, pairs or alone (Bowen, 1981; Messier & Barrette, 1982; Ozoga & Harger, 1966). This variation in social structure allows coyotes to hunt a diversity of prey species (Bowen, 1981; Fox & Papouchis, 2005). Coyotes living in groups are able to take down larger prey, such as deer. In contrast, coyotes living alone or in pairs will feed mainly on smaller mammals, such as hare and rodents. Thus, group plasticity may allow coyotes to successfully hunt in areas with different prey types. Furthermore, living as a pack allows coyotes to better defend their territory (Fox & Papouchis, 2005; Gese, 2004). As a result, they may have greater access to prey, space and potential mates (Gese, 2004). Pack living could also be the result of wolf absence in locations where coyotes and wolves formerly coexisted (Fox & Papouchis, 2005).

Without wolves, coyotes may become top predators. Forming packs may allow coyotes to fulfill this role by better enabling them to predate larger game species. Packs consist of a coyote family of up to 10 adults (Fox & Papouchis, 2005). They are hierarchal and led by the alpha male-female pair (Bowen, 1981; Fox & Papouchis, 2005). Remaining adults in a pack are called betas, which are subordinate to the alpha pair. Some betas take on the role of “helpers” and will assist with pup-rearing. Betas that do not help with pup-rearing are sometimes referred to as “slouches” (Crabtree & Sheldon, 1999b). Dispersal may occur once pups are 6 to 9 months old (Bekoff, 1977) or they may remain with the pack as future betas (Crabtree & Sheldon, 1999b). Once with the pack, coyotes may choose to leave and become more transient (Gibeau, 1993).

2.1.7 Competitors

The main competitors of coyotes are wolves (Crabtree & Sheldon, 1999b) and humans. In Grand Teton National Park, Wyoming coyote density was 33% higher in low wolf abundant areas compared to high wolf abundant areas (Berger & Gese, 2007). Furthermore, in Yellowstone National Park, coyote abundance dropped by 39% (Berger & Gese, 2007) when wolves were reintroduced in 1995 (Smith et al., 2007). Wolves actively kill coyotes, most notably around animal carcasses (Crabtree & Sheldon, 1999b; Merkle, Stahler, & Smith, 2009). Cougars will also kill coyotes to protect caches (Koehler & Hornocker, 1991). Bobcats and foxes may compete with coyotes. In contrast to wolves and cougars, coyotes are the top competitor in these interactions (Fedriani, Fuller, Sauvajot, & York, 2000). In California, coyotes killed both bobcats (Felis rufus) and gray foxes (Urocyon cinereoargenteus). As a result, gray foxes were limited to brushy habitats where coyote abundance was lower. In Maine, interspecific competition between coyotes, bobcats and red foxes (Vulpes vulpes) was researched (Major & Sherburne, 1987). Coyotes and bobcats exhibited a non-competitive relationship, however coyotes and red foxes competed over space. Coyotes also compete with swift foxes (Vulpes velox); in Texas, 89% of swift fox deaths were attributed to coyotes (Kamler, Ballard, Gilliland, & Mote, 2003).

Although competitors may influence coyote natural history, one of this study’s assumptions was that competitor presence did not affect coyote occurrence. As a result, competitors were excluded from this research.

2.2 Human-Coyote Conflict

Each year hundreds of thousands of coyotes are killed in the United States (Fox & Papouchis, 2005) and Canada (The Association for the Protection of Fur-Bearing Animals, 2013). In fact, one coyote is killed every minute in the United States (Fox, 2014); they are the most persecuted carnivore in the country (Fox & Papouchis, 2005). The belief that coyote numbers are too high and must be reduced has led to coyote killing contests (CBC News, 2015; Platt, 2016) led by individuals and organizations. In Osgood, Ontario the ‘Great Coyote Contest’ was a competition to kill the largest coyote (Alexander & Quinn, 2011). In Salmon, Idaho there is an annual ‘Coyote and Wolf Derby’ (Fox, 2012). Prizes are awarded to hunters who kill the largest wolf and the largest number of coyotes. In my opinion, coyote-related regulations, documents and programs established by some provincial governments help promote/justify human behavior towards killing coyotes. In Alberta, coyotes are considered a “nuisance” species (Government of Alberta, 2001), which means their numbers should be “controlled” (Alberta Agriculture and Forestry, 2009). Although Alberta’s Coyote Predation Control Manual and Study Guide highlights various lethal and non-lethal methods to control coyotes, the majority of the manual focuses on the lethal use of poisons (Government of Alberta, 2010). Alberta also has “relaxed” coyote hunting regulations. For example, Alberta residents do not need a license to kill coyotes on private lands so long as they have permission from landowners to hunt on their property (Alberta Environment and Parks, 2014a). I believe one of the goals of these “relaxed” hunting regulations may be to help control coyotes, as suggested by this sentence from Alberta Environment and Parks: “Alberta’s hunting community plays a vital role in wildlife management” (Alberta Environment and Parks, 2014b). Provincial governments may also sponsor coyote bounties/culls. In Saskatchewan, perceived human-coyote conflict was associated with a cull that resulted in 14

approximately 71 000 coyote deaths (Alexander & Quinn, 2011). The province paid hunters $20.00 for every coyote killed (O'Neill, 2009). Thus, provinces may sometimes encourage lethal control of coyotes. In addition to culls/bounties, recreational coyote killing contests and “relaxed” hunting regulations, coyotes are killed in the belief that lethal control will help protect livestock, game species, humans and pets (Alexander & Quinn, 2012; Fox & Papouchis, 2005). In the United States, predators accounted for 37.3% of sheep/lamb deaths in 2004 (National Agricultural Statistics Service, 2005). Other mortality causes included, but were not limited to, weather, old age, toxins, respiratory problems and disease. When considering all causes of mortality, coyotes were responsible for 22.6% of sheep/lamb deaths. In 2005, coyotes accounted for 2.4% of cattle/calf deaths in the United States (National Agricultural Statistics Service, 2006). Some of the other mortality-related causes included theft, health problems, poisoning, calving issues and weather. Predators as a whole were responsible for 4.7% of cattle/calf deaths. In an effort to decrease coyote depredation, lethal control has been used in the United States (Berger, 2006). However, it did not reduce or prevent sheep declines. Similar results have been demonstrated in South Africa (McManus, Dickman, Gaynor, Smuts, & Macdonald, 2015). There, caracals (Caracal caracal), leopards (Panthera pardus) and black-backed jackals (Canis mesomelas) were potential predators of livestock species. When ranchers in South Africa switched from lethal to non-lethal control, depredation decreased. Non-lethal control in South Africa was also less expensive than lethal control. Thus, non-lethal control of predators may be a more effective and economically sound strategy. Coyotes are also lethally removed in an attempt to increase game populations1. However, coyotes usually target the weakest individuals when hunting ungulates, for instance ungulates that are sick or starved (Crabtree & Sheldon, 1999b). Likewise, wolves tend to hunt weaker prey species (Sillero-Zubiri et al., 2004). Because of this prey preference, wolves have a smaller reproductive impact on elk populations compared to

1 Game populations are animal species that are hunted, such as deer. 15

human hunters (Wright, Peterson, Smith, & Lemke, 2006). Similar prey and hunting strategies between wolves and coyotes implies that coyotes may also affect ungulate populations less than humans. In contrast, human hunters usually kill ungulates in their reproductive prime (Milner, Nilsen, & Andreassen, 2007; Vanpe et al., 2007; Wright et al., 2006). In fact, hunter target choice is altering life history and morphological traits in prey species (Darimont et al., 2009). For example, ungulates are becoming smaller. Thus, killing coyotes to protect game species may be ineffective. Some Canadians believe coyotes should be exterminated because they pose a threat to humans (Alexander & Quinn, 2012); however, they likely exaggerate this threat as fewer than 2.7 people are bitten by coyotes each year in Canada (Alexander & Quinn, 2011). In contrast, approximately 460 000 Canadians are bitten annually by domestic dogs (Canis lupus familiaris) (Alexander & Quinn, 2011; Canada Safety Council, 2005). Furthermore, humans, rather than coyotes, are likely the root cause of human-coyote conflict. In Canada, all coyote attacks were deemed related to coyotes being food- conditioned (Alexander & Quinn, 2011). Modifying our behavior seems to be a stronger solution to reducing human-coyote conflict than extermination. To date there have been two known lethal attacks accredited to coyotes. One attack occurred in Cape Breton Highlands National Park (CBHNPC), Nova Scotia (Alexander & Quinn, 2011; National Geographic Wild). An adult female hiker, Taylor Mitchell (musician stage name), was killed on October 28, 2009 (Alexander & Quinn, 2011; National Geographic Wild). It is unclear whether the attack was by coyotes or coyote-wolf hybrids that may exist in eastern Canada (Alexander & Quinn, 2011). If the latter is the case, then it might not be reasonable to attribute the death of Taylor Mitchell to coyotes as a whole. The other lethal attack occurred in 1981 when a three-year-old girl was killed in Glendale, California (Baker & Timm, 1998). Preceding the attack, the girl’s parents had asked their neighbour to stop feeding coyotes (Baker, 2007). An investigation after the attack discovered that coyote diet in Glendale consisted of high amounts of human-associated food, such as garbage and pet food. Some Canadians are concerned about coyotes killing their pets. Between January 1995 and December 2010, 91 coyote-dog interactions were reported in the Canada print

media (Alexander & Quinn, 2011). Thirty-eight of the 91 dogs were killed; 89% of these deaths were small dogs. Particularly, 92.3% of these events involved dogs that were off- leash. As well, conflict between large and medium sized dogs appeared to be associated with coyote territorial defense. These findings suggest that leashing dogs, especially during the pup-rearing season, may help prevent coyote-dog interactions. Thirty-four coyote-cat conflicts were also reported in the Canada print media. In these interactions, 94% of cats were killed. Preventing cats from roaming by keeping them indoors or leashed may help decrease coyote-related cat mortality. Public perception of coyotes can be very negative (Alexander & Quinn, 2012). Words such as ‘vicious’ and ‘mangy’ were used to describe coyotes (Alexander & Quinn, 2012, p. 5). Additionally, many people saw coyotes as disease-ridden and a threat to their children and pets. Lethal control was seen as heroic where quotes such as, ‘euthanize’ and ‘done a kindness to thin the numbers’, were used by the press (Alexander & Quinn, 2012, p. 7). In contrast, coyote predation was seen as cruel and heartless. Statements such as ‘tore a cat to pieces’, ‘the pet died violently’ and ‘viciously attacked and savaged a cat’ were used to describe coyote-pet incidents (Alexander & Quinn, 2012, p. 6). Although coyotes have had negative interactions with people and pets, the number of these interactions have been extremely low compared to the hundreds of thousands of coyotes people kill annually (Fox & Papouchis, 2005). Coyotes have had to adjust to persecution by becoming more resilient, smarter and adaptable (Fox & Papouchis, 2005). Additionally, coyotes tend to avoid humans by shifting to more nocturnal, rather than diurnal, activity patterns (Fox & Papouchis, 2005; Gehrt et al., 2009), placing their dens in well-hidden areas, learning to evade traps and depredating livestock when humans are not around. Furthermore, coyotes may be forced to hunt livestock when there is a decrease in natural prey abundance due to human disturbance (Fox & Papouchis, 2005). As well, when humans reduce coyote numbers there is a decrease in intraspecific competition for food resources (Crabtree & Sheldon, 1999a; Fox & Papouchis, 2005). This may increase the chance of survival for remaining pups. Since more females tend to breed in persecuted areas (Fox & Papouchis, 2005) that may mean more pups are born during times when access to prey is greater. The result

may be a higher pup survival rate and an increase in coyote numbers in the long-term. The plastic social structure and reproductive ability of coyotes has allowed the species to remain resilient despite high levels of human persecution.

2.3 Ecological Importance of Coyotes

Coyotes play a vital role in the environment. They are scavengers (Crabtree & Sheldon, 1999b), they assist with seed dispersal (Young & Jackson, 1951) and as predators they help maintain biodiversity amongst prey populations (Henke & Bryant, 1999). For instance, it was demonstrated that coyote removal decreased rodent diversity by increasing mesopredator species, such as bobcats (Felis rufus), gray foxes (Urocyon cineroargenteus) and badgers (Taxidea taxa). Coyotes are also important in urban areas. They have been found to moderate deer populations in the suburbs, which can potentially limit the spread of Lyme disease (Gompper, 2002). Additionally, their predation of mesopredators, such as house cats (Felis catus), raccoons (Procyon lotor), opossums (species not specified), grey foxes (Urocyon cinereoargenteus) and striped skunks (Mephitis mephitis), helps to increase the diversity of scrub-nesting bird species (Crooks & Soulé, 1999). Moreover, they have been found to exert a regulating effect on rodents in cities like Chicago (Morey et al., 2007).

2.4 Coexistence

Because of their ecological significance and the ineffectiveness of lethal control, it is important that we coexist with coyotes rather than attempting to eradicate them. Techniques that can help promote human-coyote coexistence include practicing better livestock husbandry through the use of fences, guard dogs/llamas/donkeys and properly removing dead livestock carcasses (Fox & Papouchis, 2005). Additionally, keeping your distance from coyotes, ensuring garbage and pet food is properly stored and disposing of fallen fruit from trees are other ways to prevent potential conflict. Generally, most conflict occurs during the pup-rearing season, which means that small changes in human

behavior during critical times may improve coexistence (Lukasik & Alexander, 2011). Conflict can be especially common in off-leash areas (Alexander & Quinn, 2011). Management strategies, such as seasonal trail closures, in off-leash spaces could potentially reduce incidents during the pup-rearing season. Human-coyote conflict has been reduced through various techniques. Glendale, California, site of the first reported lethal coyote attack, used education initiatives to teach people how to coexist with coyotes (Baker & Timm, 1998). Trapping was also used. Baker and Timm (1998) argue that traps help remove “problem” coyotes as well as “[re-instill] the fear of humans in coyotes” (Baker & Timm, 1998, p. 310). In my opinion, it is unlikely that traps are effective in instilling fear of humans because coyotes may not associate traps with people. The fact that coyotes have not bitten anyone in Glendale for over 10 years may be less the product of fear of humans than the removal of “problem” coyotes through trapping (in combination, of course, with education initiatives). Another urban area that reduced human-coyote conflict was Vancouver, British Columbia. This was achieved through their Co-existing with Coyotes program, which consisted of education, ‘No-feeding and Untidy Premises Bylaws’ and aversive conditioning, such as using noisemakers and chasing coyotes (Worcester & Boelens, 2007). The combination of education and non-lethal mitigation strategies is argued to be the “only viable long- term solution to resolving conflicts with coyotes” (Fox & Papouchis, 2005, p. 2). The location of human-coyote conflict is also important to consider in management. For example, in rural settings the use of guard dogs and proper livestock carcass removal can help protect domesticated animals (Fox & Papouchis, 2005). In park areas, keeping pets on leash is important, especially during the denning season (Alexander & Quinn, 2011). Appropriate garbage disposal and removal of fallen fruit are effective ways to reduce conflict in urban areas. Thus, different locations may require different mitigation strategies. GRPP is a unique area. It is a provincial park, contains active cattle ranches and is situated between two growing urban centres. To date, there has been no human-coyote or pet-coyote research conducted there. Since location can be a factor that helps determine mitigation methods, understanding how coyotes and humans share the park may provide

insight on coexistence strategies in multi-use areas similar to GRPP that are undergoing human expansion.

Chapter Three: Comparing Coyote Occurrence Between Camera Trapping and Scat Surveys

3.1 Introduction

3.1.1 Literature Review

This research employed non-invasive methods. Unlike invasive methods, non- invasive methods do not require physical contact with the target species (Long, MacKay, Zielinski, & Ray, 2008). Capturing an animal to measure body size would be considered invasive, whereas scat surveys and camera trapping are examples of non-invasive methods (Long et al., 2008). Non-invasive methods have several advantages, especially when studying carnivore species (Long et al., 2008). Many carnivores are shy and elusive, which can make capturing efforts difficult. Consequently, exploring carnivore occurrence using non-invasive methods, such as scat surveys, may be more effective than using invasive techniques. Invasive methods may also be disadvantageous because they can be inhumane (Cattet, Boulanger, Stenhouse, Powell, & Reynolds-Hogland, 2008; Olsen, Linhart, Holmes, Dasch, & Male, 1986; Onderka, Skinner, & Todd, 1990). Traps can cause injuries, which may include swelling, lacerations and fractures (Olsen et al., 1986; Onderka et al., 1990). For example, high levels of aspartate aminotransferase (AST) and creatine kinase (CK) were found in the blood of bears trapped using leg-hold snares (Cattet et al., 2008). These enzymes “leak” into the blood from muscles when muscles have been injured. The increase of AST and CK in the bloodstream demonstrated that trapping was the probable cause of muscle damage. Capture also altered bear behavior. Their movement decreased approximately 3-6 weeks after release. Thus, invasive methods can negatively impact animals, which may support the use of non-invasive methods. Camera trapping and scat surveys are non-invasive methods used to study species occurrence (Long et al., 2008). In addition to occurrence, photos can provide information on the time an animal was present at the camera as well as animal behavior. However, cameras can be costly, may be difficult to use and may break down in the field or get

stolen. Scat surveys offer a cheap way to detect animal occurrence. They also require some expertise since it can be difficult to tell scat apart from different species (Halfpenny & Biesiot, 1986). Both scat surveys and camera trapping rely on animals leaving “signs”, such as scat and photos, to indicate their presence. If an animal occurred in an area, yet no “sign(s)” were left behind, there is the possibility of obtaining a false negative (Type II error). In this case, the researcher assumes an animal was absent when the animal was actually present. As a result, the null hypothesis is not rejected when it should be (Agresti & Franklin, 2007). Nevertheless, the relative effectiveness of scat surveys and camera trapping remains an important area of research. Studies have compared the effectiveness of camera trapping and scat surveys. Camera trapping was more effective at detecting coyote occurrence compared to scat surveys in the Grand Canyon (Reed, 2011). In contrast, research performed in the state of New York demonstrated that camera trapping was not effective at detecting coyote occurrence; scat surveys were more successful (Gompper et al., 2006). One similarity between the studies conducted by Reed (2011) and Gompper et al. (2006) was the timeframe of analysis. Research performed by Gompper et al. (2006) occurred in the summer months. Reed’s (2011) study was also primarily conducted during the summer. As well, distances between camera location and scat survey transects were fairly similar between the two studies. One aspect that differed between Gompper et al. (2006) and Reed’s (2011) research was the habitat type. Gompper et al. (2006) compared methods in “forested and anthropogenic habitats” (p. 1143), whereas Reed’s (2011) study sites were located “≥2 km from edges of developed areas” in forested locations (p. 232). Perhaps these different habitat types helped contribute to the different findings. Should vegetation be denser in the Grand Canyon, possibly from lower levels of human use, then scat may have been more difficult to find. In this case, camera trapping could potentially be the better method. Performing a methods comparison in a grassland/parkland area, such as GRPP, may provide additional insight regarding coyote detection methods. Additionally, my research compared coyote occurrence between high and low human use trails, which was not analyzed in the studies performed by Reed (2011) and Gompper et al. (2006).

Exploring method success between trail types in GRPP may strengthen animal occurrence detection for future wildlife research.

3.1.2 Significance

I compared the relative ability of scat surveys, camera trapping and a combination of these two methods to detect coyote occurrence in the grassland/parkland setting of Glenbow Ranch Provincial Park. My study differed from research performed by Gompper et al. (2006) and Reed (2011) because their studies occurred in forested areas during the summer. In contrast, my research took place in a grassland/parkland region during the spring, fall and winter. Additionally, my cameras were placed between 0 and 5 metres from trails, whereas Gompper et al. (2006) placed cameras 5 to 10 metres from bait, which was located 5 to 50 metres from trails. Reed’s (2011) cameras were not situated based on distance from trails, rather they were located in quadrats. Research performed by Gompper et al. (2006) and Reed (2011) was also different from my study because their research used baited camera stations to detect coyote occurrence. The problem with using bait is that it may be expensive, it has to be stored and disposed of properly and it may need to be continually replaced at stations (Long et al., 2008). Additionally, bait may contain human and animal-related pathogens. This is especially detrimental in GRPP since it is a well-used public park. Furthermore, baited stations may lead to bias in population demographic estimates (Mccoy, Ditchkoff, & Steury, 2011). For instance, if juveniles are more attracted to bait compared to adults, camera trapping results may reflect a population primarily consisting of juveniles, when this may not be the case. As well, bait may alter regular movement patterns by attracting animals to camera stations, which may provide inaccurate estimates of habitat use (White, Anderson, Burnham, & Otis, 1982). Because of the disadvantages associated with using bait, I used un-baited camera stations in my research.

3.2 Research Question and Hypotheses

H0: There is no significant difference between the ability of cameras, scat surveys and a combination of the two methods to detect coyote occurrence between trail types and across different seasons.

HA: Cameras, scat surveys and a combination of the two methods do not equally detect coyote occurrence between trail types and across seasons.

3.3 Study Area

Research focused specifically on high and low human use trails within GRPP. High human use trails consisted of gravel and paved paths used by park visitors. For the purpose of this research, high human use trails were connected in one 10 km loop located on the western side of the park, which contained the main access road (Glenbow Road) as well as the parking lot, park office and highest density of paved and gravel trails (Figure 3.1). The high human use trail system incorporated several trails including: Bowl Link and Badger Bowl as well as parts of Tiger Lily Loop, Yodel Loop, Glenbow Trail, Bowbend Trail and Scott Trail. I also surveyed 10 kilometres of low human use trails, which were composed of wildlife paths (Figure 3.1). Because of a large closure area in the center of the park, the 10 kilometres of low human use trails were split into a 7 km and 3 km loop. The 7 km loop was located on the western side of GRPP north of Yodel Loop. The 3 km loop was situated on the eastern side of the park and accessed by Woodland Road. All high and low human use trails were located in areas with similar features, including vegetation type and terrain.

High and lowHigh and systems trail in use GRPP. human

. 3

Figure

3.4 Methods

Scat surveys, camera trapping and a combination of these two methods (referred to as the “combined” method) were compared to understand their relative ability to detect coyote occurrence. Camera trapping and scat surveys occurred on the same trail systems to help minimize external factors confounding comparisons. For example, comparing coyote occurrence between camera traps on one trail to scat surveys on another trail may have led to bias. Seven remotely triggered cameras (Silent Image Recreational Model: RM30, Reconyx and PC900 HyperFire Professional High Output Covert IR, Reconyx) were spaced at relatively equal distances on the 10 km loop of high human use trails (Figure 3.1). Bungee cords and cables attached cameras to trees/posts that were located approximately 0 to 5 metres off the trail. Camera placement was based on obtaining high human representation. As such, many cameras were situated at trail intersections. Cameras were also placed in areas where there was a strong likelihood of capturing coyote sightings, such as near potential denning and food resource habitat. All cameras were positioned in grassland/parkland areas. Placement in similar habitat types helped reduce confounding effects. Alberta Parks had numerous cameras (Reconyx Silent Image and Hyperfire models) situated throughout GRPP on wildlife trails. Because I had placed 7 cameras on high human use trails, I selected 7 Alberta Parks cameras to represent low human use trails. Alberta Parks cameras were chosen based on their location; they needed to be in habitat types similar to high human use trail cameras to maintain consistency as well as far enough apart from one another to reduce spatial autocorrelation. Five of the low human use trail cameras were located on the 7 km loop and 2 were situated on the 3 km loop. Like the high human use trail cameras, low human use trail cameras were situated 0 to 5 metres from trails. Because Alberta Parks owned and operated low human use trail cameras, their locations are absent in Figure 3.1. The focus of the methods study was to analyze coyote trail occurrence (presence/absence on trails) as opposed to determining coyote abundance/density. This helped decrease error associated with spatial autocorrelation. For instance, if occurrence 26

was monitored based on abundance/density, then a coyote caught at two cameras could be interpreted as two different coyotes, which would overestimate coyote numbers. In this case, to prevent double counting, coyotes would need to be recognized at the individual level. Due to time constraints I was not able to identify different coyotes. Consequently, I monitored occurrence based on presence/absence to help reduce pseudoreplication. If two separate cameras detected coyote occurrence, it would simply demonstrate coyote trail use at both camera locations. Cameras were active between June 2014 and June 2015. However, many of the cameras in the park failed to collect data for the entire timeframe. Reasons included dead batteries, full camera cards, cold temperatures and public tampering. As a result, the methods study was split into two time periods. The first period was between September 11, 2014 and November 26, 2014 (Time Period 1). During Time Period 1 I compared methods by trail type because it was the only time period when all 14 cameras were functional on both high and low human use trails. The second time period was between September 11, 2014 and June 16, 2015 (Time Period 2), which investigated methods across different seasons. Seasonal categories included: fall (September 23 to December 20), winter (December 21 to March 19) and spring (March 20 to June 20). This analysis was restricted to the 7 cameras located on high human use trails because, unlike the low human use trail cameras, they functioned consistently throughout Time Period 2. Cameras remained stationary for the duration of the study. The 7 cameras on the high human use trails were checked once every two weeks (biweekly) to change memory cards and batteries (Brichieri-Colombi, 2012). If road conditions were poor or the temperature was lower than -20°C, fieldwork was moved to another day for safety reasons. Cameras owned by Alberta Parks were monitored based on Alberta Parks volunteer availability. I brought memory cards obtained from high human use trail cameras back to the lab where photos were uploaded to a computer. Photos were coded based on coyote occurrence using MapView Professional software (version 3.4.0.15 © Reconyx, Inc. 2005-2014). From MapView, data was exported to Microsoft Excel 2010. Alberta Parks volunteers collected memory cards from low human use trail cameras and coded the

pictures. These volunteers were taught classification protocol by an ecologist in species identification and by the Assistant to Operations Manager for Alberta Parks Kananaskis. The Operations Manager error-checked the data before emailing it to me in Excel format. I combined the high and low human use trail data in Excel. Once combined, I created the trail type variable. Dates associated with exported photograph data allowed me to create the season variable. Excel data was then transferred to TIBCO Spotfire S+® (version 8.2, © 1988, 2010 TIBCO Software Inc.) for statistical analysis. Concurrent with camera work, high and low human use trails were walked once every two weeks to perform scat surveys. Coyote scat is distinct from scat of other species found in the park (Appendix A). For example, the section of feces that leaves the coyote’s anus is pointed (Halfpenny & Biesiot, 1986). In comparison to other canid scat, coyote scat diameter is approximately 18 to 25 millimetres, whereas fox scat diameter is usually less than 18 millimetres. Coyote feces may consist of bone fragments and animal hair. In contrast, domestic dog scat often appears “grainy” (Lukasik & Alexander, 2012). When it snowed in GRPP, I waited at least one day after snowfall to ensure scat visibility. Field excursions were reassigned to another day if road conditions were poor and/or if the temperature was colder than -20°C. I searched for scat located on or one metre off the trail. For each two-week period, I recorded coyote occurrence based on the presence of scat. For example, if I found scat on low human use trails, I recorded coyote occurrence as present during that specific two-week period. If I did not find scat on low human use trails, then coyote occurrence was recorded as absent over the two-week period. This protocol also applied to high human use trails. Like scat surveys, coyote occurrence was recorded as present/absent for each two-week period on each trail type for camera trapping and the “combined” method. When scat and photo data were pooled for the “combined” method, coyote occurrence was not double-counted. For example, if scat surveys and camera trapping detected coyote occurrence during the same two-week period it was considered as one overall occurrence. Scat was collected and stored in a freezer at the University of Calgary for a future diet study. Scat UTM coordinates, obtained using a GPS (Garmin GPSMAP® 62s and Garmin eTrex Legend H), were also recorded for this future study. In the lab, UTM

coordinates were uploaded to GPS Utility (version 5.20, © GPS Utility Ltd. 1998-2011), transferred to Microsoft Excel 2010 and inputted into ArcMap (Esri ® ArcMap ™ 10.1, © 1999-2012 Esri Inc.) for future visual analysis. ArcMap helped create the maps depicted in Figures 1.1 and 3.1. The park boundary layer (Figure 1.1) and the closure area layer (Figure 3.1) were provided by Alberta Tourism, Parks and Recreation (2013). The datum and projection used for Figure 1.1 was NAD 1983 3TM 114, which was the original projection of the layers in this figure. The 3TM (3-degree Transverse Mercator) is a modified projection. In Alberta, 3TM is generally used for cities and towns (Alberta Sustainable Resource Development, 2005). It is 3 degrees wide compared to UTM zones, which are 6 degrees wide. In the case of the 3TM projection used in Figure 1.1, the central meridian is located at 114 degrees. In contrast, the datum and projection of Figure 3.1 was WGS 1984 UTM Zone 11N. Layers were not re-projected to NAD 1983 3TM 114 because most of the layers were originally projected as WGS 1984 UTM Zone 11N. Because there is error associated with re-projecting map layers (Esri ® ArcMap ™ 10.1, © 1999-2012 Esri Inc.), datum WGS 1984 and projection UTM Zone 11N were used. Any layers in Figure 3.1 that did not have the same datum and projection were re-projected to WGS 1984 UTM Zone 11N to maintain consistency (Data Resource Centre, 2012).

3.4.1 Statistical Justification and Analysis

Coyote occurrence values during Time Period 1 were the same for all method and trail types. As a result, no statistical analysis was necessary. In contrast, coyote occurrence values obtained during Time Period 2 differed by method and season. For this analysis I performed a Fisher’s exact test. The Fisher’s exact test was ideal because it can handle small sample sizes (McDonald, 2014). Furthermore, my data failed the assumptions of other statistical tests that were initially considered. For example, the dependent variable (coyote occurrence)

was nominal1, as opposed to ordinal, which failed an important assumption of the Kruskal Wallis test (Siegel, 1956). Additionally, the Kruskal Wallis test ranks values. Because some of the season and method type values were identical, I believed that ranking tied values would detract from the integrity of the data. I had also considered the two-way ANOVA, however the dependent variable was not continuous (Laerd Statistics, 2013c). Because the data set did not meet assumptions of other statistical tests, the Fisher’s exact test was ideal. Furthermore, the Fisher’s exact test has been used in other ecological studies (e.g. Cook, Awmack, Murray, & Williams, 2003; Peterson & Robins, 2003; Sullivan, 1988). Significance was determined using an alpha value of 0.05.

3.5 Results

Coyote occurrence was detected over all two-week periods during Time Period 1 (Table 3.1). Consequently, statistical analysis was not necessary. There was no difference in coyote occurrence between method and trail types.

Table 3.1 Number of two-week periods between September 11, 2014 and November 26, 2014 that detected coyote occurrence based on method and trail type. Trail Type Method Type # Two-Week Periods that Detected Coyote Occurrence High Human Use Scat Survey 5 High Human Use Camera Trapping 5 High Human Use Combined Method 5

Low Human Use Scat Survey 5 Low Human Use Camera Trapping 5 Low Human Use Combined Method 5

During Time Period 2, there was no statistical difference between method types during the spring (Fisher’s exact test, p = 0.30), fall (Fisher’s exact test, p = 1.00),

1 Nominal and ordinal data are considered categorical for the purpose of this research, which is what SPSS follows. 30

winter (Fisher’s exact test, p = 1.00) and combined seasons (Fisher’s exact test, p = 0.32) (Table 3.2).

Table 3.2 Number of two-week periods between September 11, 2014 and June 16, 2015 that detected coyote occurrence using various method types across different seasons. Season Method Type # Two-Week Periods that Detected Coyote Occurrence Spring Scat Survey 5 Camera Trapping 7 Combined Method 7

Fall Scat Survey 7 Camera Trapping 7 Combined Method 7

Winter Scat Survey 6 Camera Trapping 6 Combined Method 6

Combined Scat Survey 18 Seasons Camera Trapping 20 Combined Method 20

3.6 Discussion

Scat surveys, camera trapping and the “combined” method detected coyote occurrence equally on both trail types. This result was surprising in the case of low human use trails; I expected that camera trapping would detect coyote occurrence better than scat surveys. Characteristics of low human use trails, such as narrow trail width and dense vegetation, made scat challenging to find. Fallen leaves and snow also increased the difficulty of detecting scat. In contrast, leaves and snow did not prevent cameras from taking pictures, so long as the sensor was not covered. Moreover, precipitation rarely made objects/animals too “blurry” to identify in photos. Performing scat surveys on low human use trails was also challenging because snow and dense vegetation sometimes made the trail difficult to find. Although I used a GPS to follow trail coordinates, the GPS 31

resolution was ≤ 3 metres. As a result, I may have walked up to 3 metres off the trail in areas where the trail was not visible, potentially overlooking coyote scat. In contrast, cameras consistently monitored coyotes in the same location over the duration of the study. Scat surveys were also difficult when low human use trails expanded into wildlife corridors. Wildlife corridors are areas animals can traverse through as opposed to traveling on a specific trail. Thus, coyotes may have traveled through sections of the wildlife corridor that did not include the low human use trail, which could have decreased the likelihood of finding scat. Conversely, cameras detected coyote occurrence over a wider area of the corridor compared to scat surveys. In ideal conditions, RM30 and PC900 cameras could photograph animals approximately 30 metres away from the camera location (Reconyx, 2006, 2013). Despite the difficulty of finding scat on low human use trails, all methods detected coyote occurrence equally. Thus, there was either enough scat on low human use trails to allow for coyote occurrence detection, or low human use trail characteristics did not affect the ability to detect scat. There was no statistically significant difference between method types during the fall, winter and spring on high human use trails (Fisher’s exact test, p = 1.00; Fisher’s exact test, p = 1.00; Fisher’s exact test, p = 0.30). This was also unanticipated. I expected scat surveys would better detect coyote occurrence on high human use trails during the fall and spring. Scat was easily seen on high human use trails because the trails were wide and relatively flat. Additionally, vegetation alongside the trails was often sparse or mowed by Parks staff, which helped increase scat visibility. As well, I found large numbers of scat on high human use trails when I scouted the study area a month before research began. The fact that camera trapping and scat surveys equally detected coyote occurrence on high human use trails during the spring and fall may have been a result of effective camera placement. Cameras were located in areas where coyote occurrence was expected, such as near good denning habitat. Additionally, cameras were spaced far enough apart to reduce spatial autocorrelation, while close enough to ensure coyote detectability throughout the study area (Silver, 2004). Over the winter, I believed that camera trapping would detect coyote occurrence better than scat surveys because snow made scat detection difficult. Additionally, cameras on high human use trails

functioned during the coldest days of the year. During these cold periods, I had to postpone scat surveys to warmer days of the week. The ability of cameras to continue functioning during low temperatures when I was unable to perform scat surveys demonstrated that camera trapping could potentially be the better method to detect coyote occurrence in the winter. Regardless, all methods detected coyote occurrence equally on high human use trails during this season. Although there was no statistically significant difference between method ability to detect coyotes across the seasons, scat surveys did not detect coyote occurrence during 2 two-week periods. Specifically, the first two-week period was from May 8, 2015 to May 19, 2015 and the second two-week period was from June 2, 2015 to June 16, 2015. During these periods field notes indicated that there was no snow on trails. Consequently, scat would have been easily detected without snow cover. Why did camera trapping detect coyote occurrence during these two-week periods when scat surveys did not? Did coyote trail use shift? Did coyotes use the trails, but not defecate on the trails? If so, could this be associated with a shift in energy allocation between marking territory and taking care of pups in the spring? Future research could explore if this result consistently occurs during spring, and if so, whether a decrease in defecation on high human use trails could be related to coyote natural history. Aside from detecting occurrence, there may be additional features of camera trapping and scat surveys that researchers may want to consider. Because I did not need to monitor cameras as often as I needed to perform scat surveys, cameras may be better suited for study areas in remote locations or areas that are difficult to traverse. Researchers may also want to determine if they need to collect additional information apart from occurrence. For example, photos provide a wealth of information, such as temperature, moon phase, activity type and time. Videos can also be recorded on cameras, which may help understand animal behavior (Long et al., 2008). On the other hand, scats can be dissected to explore animal diet and DNA samples collected from scat can help identify individuals. Cost is another important feature to consider when comparing camera trapping and scat surveys. Cameras are expensive, and may require additional purchases, such as batteries, memory cards, cables and locks. It is also

important to buy quality cameras, which may increase costs. I found that pictures taken from older cameras were not as sharp, which sometimes made species identification difficult. Additionally, some of the older cameras were not as waterproofed, had troubles reading memory cards and stopped working once the temperature became too cold. Cameras may also get stolen, as was the case with one of my cameras at the end of the study, which could add to research expenses. In contrast, scat surveys are a cheaper way to detect animal occurrence. The amount of time needed to learn a method may be another important factor to take into account. For example, I had to learn how to identify coyote scat from scat of other species. I also had to learn how to use cameras and trouble- shoot them in the field. Consequently, previous knowledge may influence the type of method chosen for a study. Thus, there are additional factors that may need to be considered when selecting an appropriate method.

3.7 Conclusion

This study helped shed light on non-invasive methods used for identifying coyote occurrence in a grassland/parkland region. Unlike research performed by Reed (2011) and Gompper et al. (2006), there was no statistically significant difference between camera trapping and scat surveys in GRPP. These contrasting results may have been due to different timeframes of analysis and habitat types. Research performed in GRPP occurred during the spring, fall and winter compared to the studies conducted by Reed (2011) and Gompper et al. (2006), which occurred primarily in the summer. Additionally, the study areas used by Reed (2011) and Gompper et al. (2006) were more forested than GRPP. Future research could further explore how seasonality and characteristics of research sites affect method success.

Chapter Four: Coyote Occurrence Relative to People in GRPP

4.1 Introduction

4.1.1 Literature Review

Studies have documented coyote spatio-temporal occurrence relative to people. In Cape Breton Highlands National Park of Canada (CBHNPC), Nova Scotia, coyote activity decreased as cyclist and hiker activity increased during the summer and fall (Porter, 2013). Human and coyote activity were not correlated during the winter. In a study area west of Los Angeles, California, coyotes selected for natural spaces compared to areas that were inhabited by people (Riley et al., 2003). Similarly, research conducted on the Nature Reserve of Orange County (NROC), California, demonstrated that coyote occurrence was significantly lower in high human use areas compared to low human use areas (George & Crooks, 2006). Coyote occurrence may also vary temporally in relation to people. For instance, coyote nocturnal activity increased in locations with high levels of human presence (Gehrt et al., 2009; Kitchen, Gese, & Schauster, 2000; Riley et al., 2003). Thus, humans may affect coyote occurrence both at the spatial and temporal levels.

4.1.2 Significance

I examined how coyote trail occurrence changes seasonally, daily and throughout stages of the coyote’s life cycle relative to human presence in GRPP. This study occurred between July 2, 2014 and January 23, 2015 inclusive. My work was similar to Porter’s (2013), however her study excluded November and December. Because November and December are important dispersal months for coyotes, it would be interesting to see patterns of coyote occurrence on trail systems during these times. Furthermore, Porter’s (2013) analysis was restricted to potential seasonal changes in coyote occurrence as opposed to daily and lifecycle changes. Like my research, the studies performed by Riley et al. (2003) and George and Crooks (2006) also occurred in natural areas that had varying types of human activities. However, Riley et al. (2003) and George and Crooks (2006) observed daily changes in coyote occurrence as opposed to seasonal and lifecycle

changes. I believe it is important to explore coyote occurrence across various timeframes. This type of in-depth analysis may provide a more detailed summary of coyote trail occurrence, which may help offer insight as to why trail use patterns exist.

4.2 Research Question, Objectives and Hypotheses

Research Question 2: How do coyotes and humans co-occur spatio-temporally within GRPP on high and low human use trails? This question was broken into two objectives:

H0: Coyote occurrence does not differ by trail type, temporal timeframe or between trail types across various timeframes.

HA: Coyote occurrence varies depending on the trail type, temporal timeframe as well as between trail types across various timeframes.

Research Objective 2: How do human-related disturbances, past coyote trail occurrence, prey, moon phase, season, life cycle stage and time of day relate to coyote occurrence on trails?

H0: Coyote occurrence is not affected by human-related disturbances, past coyote trail occurrence, prey, moon phases, seasons, life cycle stages and times of the day.

HA: Coyote occurrence is affected by human-related disturbances, past coyote trail occurrence, prey, moon phases, seasons, life cycle stages and times of the day.

4.3 Study Area

I used five new low human use camera locations for this study. These camera locations are absent in Figure 3.1 to respect their ownership by Alberta Parks.

4.4 Methods

Camera methods used in this chapter followed the same methods outlined in Chapter 3 (p. 21). Method details specific to Chapter 4 are outlined below. The dependent variable for Objective 1 and Objective 2 was coyote trail occurrence detected by cameras. Specifically, for Objective 1, coyote occurrence was tallied per week. Although counting coyotes can potentially lead to pseudoreplication, as described in section 3.4 (p. 26), I chose to tally coyote numbers for Objective 1 because I felt that the statistic that best addressed my hypothesis was a two-way ANOVA (refer to section 4.4.1.1, p. 41). Because the two-way ANOVA requires a continuous dependent variable, I needed to count coyote occurrence, rather than measure occurrence based on coyote presence/absence. To minimize pseudoreplication, I used a Time to Independence (TTI) value specific to coyotes, which was determined by Swihart, Slade and Bergstrom (1988). TTI represents the amount of time needed between occurrence observations for temporal autocorrelation to be negligible; it signifies “the time interval at which an animal’s current position was influenced only by its pattern of home range use, not by its position Δt minutes earlier” (Swihart, Slade, & Bergstrom, 1988, p. 394). TTI has been used in other studies to reduce double counting (e.g. MacDonald, 2012; Smith, 2009). Swihart, Slade and Bergstrom (1988) determined that the amount of time it took a coyote to traverse a home range of 6.185 km2 averaged 207 minutes, which is approximately 3 ½ hours. Because coyote home range in GRPP is likely between 5 to 8 km2, as mentioned in section 2.1.3 (p. 9), I waited 3 ½ hours before tallying could resume after a coyote was observed. For Objective 2, coyotes were considered present or absent. Coyote occurrence was binary so that I could perform a binary multiple logistic regression (BMLR), as described in section 4.4.1.2 (p. 43). Data was amalgamated by time of day, which

included morning twilight, daytime, evening twilight and nighttime. During the spring and summer months, when days grew longer, time categories included: morning twilight (4:00 am - 5:00 am), daytime (5:00 am - 8:00 pm), evening twilight (8:00 pm - 10:00 pm) and nighttime (10:00 pm - 4:00 am) (National Research Council Canada, 2015). During the fall and winter, when days grew shorter, time categories included: morning twilight (6:00 am - 8:00 am), daytime (8:00 am - 6:00 pm), evening twilight (6:00 pm - 7:00 pm) and nighttime (7:00 pm - 6:00 am) (National Research Council Canada, 2015). Time ranges were based on times experienced in Calgary, Alberta. Independent variables for Objective 1 were trail type (high and low human use trails) as well as three temporal variables. The latter included: season (summer, fall, winter), life cycle stage (weaning, dispersal) and time of day (morning twilight, daytime, evening twilight, nighttime). Dates and times associated with photographs allowed me to create these temporal variables. Summer was categorized between June 21 and September 22. Fall was from September 23 to December 20 and winter was between December 21 and March 19. Life cycle categories were based on various suggestions in the literature (Lukasik & Alexander, 2012; Tesky, 1995). The weaning season, which is near the end of the pup-rearing season, was defined as May 15 to August 31. The dispersal season was defined as September 1 to January 31. Time of day categories were described earlier in this section. There were 10 independent variables in Objective 2 (Table 4.1). Like the dependent variable, independent variables were amalgamated by time of day. Data for the 5 ratio independent variables (Number of People, Number of Prey, Number of Vehicles, Number of Cyclists and Number of Dogs) was obtained when photos were initially coded using MapView Professional software, as described in section 3.4 (p. 26). Specifically, the Number of People variable was based on the total number of people using trails including, but not limited to, cyclists, pedestrians, skateboarders and rollerbladers. I did not count people in vehicles because I could not differentiate individuals in all vehicle types based on photographs. Furthermore, because people were often concealed within vehicles, I believed that coyotes might not associate vehicles with humans. The Number of Prey variable represented a count of mule deer (Odocoileus hemionus), white-tailed

deer (Odocoileus virginianus), elk (Cervus canadensis), moose (Alces alces), ungulates that could not be identified to species, black-billed magpies (Pica hudsonia), snowshoe hares (Lepus americanus), American crows (Corvus brachyrhynchos), common ravens (Corvus corax), grouse species, white-tailed jackrabbits (Lepus townsendii), red squirrels (Tamiasciurus hudsonicus) and Richardson’s ground squirrels (Urocitellus richardsonii). Any type of motorized vehicle was also counted (Number of Vehicles variable). This included, but was not limited to, vans and trucks (e.g. Alberta Parks vehicles and Conservation Officer trucks), golf carts, quads and tractors. The Number of Cyclists variable represented the number of people riding bicycles on trails. The Number of Dogs variable represented the number of domestic dogs, both un-leashed and leashed, on trails.

Table 4.1 Variables used in the binary multiple logistic regression. Variable Dependent/Independent Data Type Variable Groups

Coyote Occurrence Dependent Nominal Absence Presence Season Independent Nominal Summer Fall Winter Life Cycle Stage Independent Nominal Weaning Dispersal Time of Day Independent Nominal Morning Twilight Daytime Evening Twilight Nighttime Coyote Occurrence in Independent Nominal Presence Past Day Absence Moon Phase Independent Ordinal New moon Crescent moon Quarter moon Gibbous moon Full moon Number of Cyclists Independent Ratio Count data Number of People Independent Ratio Count data Number of Dogs Independent Ratio Count data Number of Prey Independent Ratio Count data Number of Vehicles Independent Ratio Count data

Categorical variable data for Objective 2 was obtained in various ways. As described earlier in this section, season, life cycle stage and time of day variables were

created based on date and time data associated with photos. These temporal variables followed the same categorization as described previously. Moon phase data was provided with each high human use trail photo and exported to the Excel dataset. However, moon phase was not included with Alberta Parks data (low human use trail cameras). As a result, I manually added moon phases associated with Calgary, Alberta to the Alberta Parks dataset (The Old Farmer's Almanac, 2015). Another categorical independent variable was coyote occurrence on trail systems within the past 24 hours (Coyote Occurrence in Past Day). This variable was created using data from the dependent variable (Coyote Occurrence). For example, the Coyote Occurrence value for November 2 became the Coyote Occurrence in Past Day value for November 3. Camera trapping, not scat surveys, was used for both Objectives. Data required for Objective 2 could not be obtained from scat, such as moon phase and the number of people. Scat surveys were not included in Objective 1 because not all cameras were operational on trail systems associated with scat surveys. In order to use camera trapping and scat surveys, they both needed to occur on the same trails. I could not measure coyote occurrence using camera trapping on one trail while performing scat surveys on another trail because it could lead to bias. The problem was that the 7 cameras associated with the 10 kilometres of surveyed low human use trails were only consistently functional between September 11, 2014 and November 26, 2014. This reduced timeframe was detrimental because it would not have allowed for a seasonal and life cycle stage analysis. One option was to move the analysis to exclusively high human use trails where cameras were more functional. This would allow for a longer study that could incorporate different temporal analyses. Additionally, coyote presence could be determined using camera trapping as well as scat surveys since both methods were used on the 10 km loop of high human use trails. However, I would not be able to compare occurrence between trail types, which was an important aspect of the study. Because scat surveys did not improve coyote occurrence detection, as demonstrated in Chapter 3, I decided to solely use camera data. This allowed me to select the top 7 Alberta Parks cameras that consistently collected the most amount of data on low human use trails, regardless of whether the trails were associated with scat surveys. Five of these cameras were different,

and therefore in different locations, from the ones used in Chapter 3. These 5 new cameras were located in similar habitat types as the other cameras used in the study. The timeframe for both Objectives was between July 2, 2014 and January 23, 2015.

4.4.1 Statistical Justification and Analysis

Both research objectives used SPSS (SPSS Statistics, version 22, 64-bit edition, © IBM Corporation 1989, 2013) software. Different statistical analyses were required for Objectives 1 and 2. They are described separately in sections 4.4.1.1 and 4.4.1.2 below.

4.4.1.1 Research Objective 1

I used the two-way ANOVA test to analyze Objective 1. Although the data was non-parametric, the two-way ANOVA is fairly robust with non-normally distributed data (Laerd Statistics, 2013c). This test is commonly used in ecology (e.g. Clegg & Owens, 2002; Petchey, McPhearson, Casey, & Morin, 1999; Silva, Uhl, & Murray, 1996) and is an effective method for comparing means between two independent variables (Laerd Statistics, 2013c). For my research, the two-way ANOVA analyzed coyote occurrence relative to each independent variable individually, known as a within comparison. For example, is season significantly related to coyote occurrence? The two-way ANOVA also examined coyote occurrence relative to two independent variables together, known as the between comparison. For example, are both season and trail type significantly related to coyote occurrence? Because there were three temporal variables, I performed three two- way ANOVA tests: 1) coyote occurrence relative to trail type and season 2) coyote occurrence relative to trail type and life cycle stage 3) coyote occurrence relative to trail type and time of day.

The dataset fit the two-way ANOVA test because the dependent variable was continuous, the independent variables were categorical and the observations were independent (Laerd Statistics, 2013c). Additionally, the impact of significant outliers on continuous variables was reduced through winsorization (Howell, 2014). Winsorizing a dataset is when outlier values are replaced with the highest and lowest values of data that

are not considered outliers (Figure 4.1). In order to identify outliers, values were first converted to z-scores (Field, Miles, & Field, 2012). Original values associated with z-scores > ±3.29 were winsorized.

A. 1 1 2 9 10 10 10 11 13 14 20 20 21

B. 9 9 9 9 10 10 10 11 13 14 14 14 14

Figure 4.1 Winsorization process. The lowest outliers in A are 1, 1 and 2. The highest outliers are 20, 20 and 21. B represents the winsorized data. The lowest value that is not an outlier, which is 9, replaces the 3 lower outliers. The highest value that is not an outlier, which is 14, replaces the 3 higher outliers.

Based on significant Levene’s statistics, the two-way ANOVA assumption of homoscedasticity was violated (season: F(5,54) = 4.87, p = 0.001, life cycle stage: F(3,56) = 6.76, p = 0.001, time of day: F(7,232) = 25.048, p < 0.001) (Laerd Statistics, 2013c). This was tolerable for the analysis that explored coyote occurrence between trail types during periods of the day because the variable group1 population sizes were equal (Zar, 1999). However, group population sizes were unequal in the seasonal and life cycle stage variables. Thus, the alpha value for these two tests was decreased from 0.05 to 0.01 to help prevent Type I error (false positive) (Sarstedt & Mooi, 2014). The alpha value for the time of day analysis remained at 0.05. The two-way ANOVA was not able to perform simple effect testing. This is when independent variable groups are analyzed relative to coyote occurrence (IBM, 2014; Weinberg & Abramowitz, 2008). For example, is coyote occurrence different between high and low human use trails during the summer? In a way, simple effect testing is similar to performing multiple t-tests between variable groups. The problem with

1 Variable groups are the categories within a variable. For example, morning twilight, evening twilight, daytime and nighttime are groups within the time of day variable. 42

performing multiple t-tests is the increased chance of Type I error (Laerd Statistics, 2013b). Consequently, I used the Bonferroni test to perform simple effect testing, which helps control for this (Field, 2009; Laerd Statistics, 2013b).

4.4.1.2 Research Objective 2

The goal was to explore how the 10 independent variables (Table 4.1) were related to coyote occurrence. A regression analysis was ideal because it would provide information about which variables were the most important predictors of occurrence (ReStore, 2011). I did not use multiple linear regression because three of the independent nominal variables had multiple groups, as opposed to being either “continuous or dichotomous” (Tabachnick & Fidell, 1996, p. 128). Furthermore, continuous variables were not normally distributed (Gould & Gould, 2002) even after Logarithm base 10, square root, inverse and Natural Logarithm transformations (Table 4.2).

Table 4.2 Skewness and kurtosis values of the data.

Variable No Logarithm Square Root Inverse Natural Transformation Base 10 Logarithm Skew Kurt Skew Kurt Skew Kurt Skew Kurt Skew Kurt # Cyclists 3.89 15.44 2.00 3.01 2.37 5.21 -1.29 -0.13 2.00 3.01 # People 3.32 11.85 0.66 -0.72 1.62 2.41 0.31 -1.70 0.66 -0.72 # Dogs 3.06 9.69 1.42 0.97 1.54 1.68 -0.74 -1.18 1.42 0.97 # Prey 2.21 4.74 1.24 0.36 1.09 -0.76 -0.79 -1.10 1.24 0.36 # Vehicles 2.83 7.40 2.01 2.79 1.94 2.50 -1.60 0.81 2.01 2.79 Note. Skewness (Skew) and kurtosis (Kurt) values were calculated before transformations as well as after the Logarithm Base 10, Square Root, Inverse and Natural Logarithm transformations were applied. All standard error values for skewness were 0.045 and all kurtosis standard error values were 0.091. When the skewness value is divided by its standard error the data is considered non-parametric if the resulting value is > ±1.96 (Rose, Spinks, & Canhoto, 2015). All skewness values were over ±1.96. Thus, the transformations did not have normal distributions.

Instead of multiple linear regression, I used binary multiple logistic regression (BMLR) for Objective 2. Logistic regression is a common statistic used to compare species occurrence against multiple independent variables (e.g. Crooks, Suarez, Bolger, & Soulé, 2001; George & Crooks, 2006; Niedzialkowska et al., 2006; Ordeñana et al.,

2010; Tynan et al., 2005). In BMLR the dependent variable is binary and there is more than one independent variable (McDonald, 2014). This is in contrast to multinomial multiple logistic regression where the dependent variable has more than 2 groups. BMLR was appropriate for my data because it is tolerable of non-parametric data (Tabachnick & Fidell, 1996). As well, independent variables can be categorical and/or continuous. There are several assumptions associated with BMLR. One assumption is that there are over 10-20 rows of data per independent variable (Williams, 2012). My dataset met this assumption. Secondly, the logit of the dependent variable must be linearly related with each continuous independent variable (Laerd Statistics, 2013a). This was analyzed using the Box-Tidwell transformation test (Williams, 2012). Results demonstrated that 4 of the 5 continuous independent variables were not linearly related to the logit. However, this assumption may be violated because “it appears that almost whenever a continuous variable is a significant predictor it also violates linearity of the logit” (Williams, 2012, p. 97). Lack of correlation between independent variables is another assumption (Williams, 2012). This was explored using a Spearman’s rho test. Independent variable pairs that had statistically significant values greater than ± 0.80 were considered too highly correlated (Simkiss, Edmond, Bose, Troy, & Bassat, 2015). This was the case for 2 variable sets. Consequently, one variable from each pair was removed from the analysis (Williams, 2012). Table 4.1 highlights the final variables used in the BMLR. Lastly, the impact of univariate outliers associated with continuous independent variables was decreased through winsorization (Field et al., 2012; Howell, 2014). Multivariate outliers were identified using the Cook’s Distance test (Williams, 2012); three rows of data were associated with multivariate outliers and were deleted from the dataset. I first performed a forward stepwise BMLR. Variables that best predicted coyote occurrence (i.e. p-values < 0.05) were added to the model one at a time (McDonald, 2014; Nau, 2015). Because I used the forward stepwise approach as an exploratory process, I believed that having a higher number of variables in the model would help broaden my understanding of how different variables were related to coyote occurrence as well as what variables were the strongest predictors of occurrence. As such, variables

with p-values > 0.20 were excluded from the model; this high p-value decreased the chance that variables would be rejected. The resulting equation from the stepwise regression included all variables, except Moon Phase and the Number of Dogs. After the forward stepwise regression, I created BMLR models and used Akaike Information Criteria (AIC) to compare these models (Equation 1) (Burnham & Anderson, 2002). AIC ranks models using two key factors. The first deals with the principle of parsimony; the fewer the variables (k) in a model the better the model is. The second factor is based on how well the model fits the “full truth or reality” (Burnham & Anderson, 2002, p. 96). This is depicted as the loglikelihood (goodness of fit) in Equation 1. Once the AIC is obtained for each model, AICc can be calculated, which helps account for smaller sample sizes (Burnham & Anderson, 2002). The lowest AICc value is then subtracted from all model AICc values to obtain the delta AIC (Δi). Delta AIC values can be used to predict the best model. A value between 0 and 2 demonstrates “substantial” support for the model (Burnham & Anderson, 2002, p. 70). If the value is between 4 and 7 there is “considerably less” support, while values greater than 10 demonstrate “essentially [no]” support for the model. The model with the lowest Δi value is considered the top model.

AIC = -2loglikelihood + 2K (1)

Delta AIC values can also be used to determine Akaike weights (wi) for each model (Burnham & Anderson, 2002). Dividing two wi values of different models provides a ratio value, which allows for a quantitative comparison. For example, if Model 1 has a weight of 3 and Model 2 has a weight of 1, then Model 1 has three times the explanatory power of Model 2. Models 1 to 9 were the first models created (Table 4.3). Variables were selected based on coyote natural history and biological reasoning (Appendix B). These first nine models were compared using AIC; the top model was Model 1 (Global) and the second best model was Model 2 (Natural Factors). The Natural Factors model lacked variables that were human-related disturbances (i.e. number of people/dogs/cyclists/vehicles). In 45

contrast, the Global model included all variables. I wanted to know if the addition of specific human-related variables to the Natural Factors model would create a stronger model than the Global model. As a result, models 10 to 14 were created. All 14 models were then compared using AIC. Resulting beta coefficients of the top model helped explain the relationship between the dependent and independent variables. P-values < 0.05 were deemed significant.

Table 4.3 Models created for binary multiple logistic regression. Model Number and Name Independent Variables 1. Global All variables 2. Natural Factors Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase, Number of Prey 3. Human-Related Disturbances Number of Cyclists, Number of People, Number of Dogs, Number of Vehicles 4. Time-Related Season, Life Cycle Stage, Time of Day 5. Animals Coyote Occurrence in Past Day, Number of Dogs, Number of Prey 6. Coyote Social Aspects Coyote Occurrence in Past Day, Moon Phase 7. People and Dogs Number of People, Number of Dogs 8. People Number of People 9. Moon Phase Moon Phase 10. Natural Factors + Cyclists Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase, Number of Prey, Number of Cyclists 11. Natural Factors + Vehicles Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase, Number of Prey, Number of Vehicles 12. Natural Factors + Cyclists + Vehicles Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase, Number of Prey, Number of Cyclists, Number of Vehicles 13. Natural Factors + People Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase, Number of Prey, Number of People 14. Natural Factors + Dogs Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase, Number of Prey, Number of Dogs

4.5 Results

4.5.1 Research Objective 1: Is coyote occurrence less on high human use trails compared to low human use trails? Does coyote trail occurrence fluctuate across different seasons, life cycle stages and times of the day?

The first two-way ANOVA analysis explored coyote occurrence relative to season and trail type. Regarding the within comparison, season was significantly related to occurrence, F(2, 54) = 19.796, p <0.001, while trail type was not, F(1, 54) = 2.074, p = 0.156. Simple effect testing demonstrated that more coyotes occurred on trails during the fall and winter compared to the summer based on harmonized mean values (Figure 4.2).

10 * 8 * 6 Summer Fall 4 Winter

2 Average Coyote Abundance/Week

Summer/Fall Summer /Winter Fall/Winter

Figure 4.2 Comparison of harmonized mean coyote occurrence per week between summer, fall and winter on all trail types. Asterisks indicate significance levels when p < 0.01. Error bars represent standard error.

The two-way ANOVA comparison between season and trail type was not statistically significant, F(2, 54) = 0.839, p = 0.438. Thus, coyote occurrence during different seasons did not vary by trail type. Likewise, occurrence on trail type did not vary by season. Simple effect testing demonstrated that coyote occurrence on high and low human use trails was not significantly different during the summer, fall and winter

(Table 4.4). However, there was a lower p-value of 0.124 when comparing trail type in the winter. During this season, harmonized mean values indicated that there were 10.4 (SE = 1.6) coyotes on high human use trails per week versus 6.8 (SE = 1.6) coyotes on low human use trails per week.

Table 4.4 Comparison of coyote occurrence between trail types within each season. Season Trail Type P-value Summer High Human Use vs. Low Human Use 0.956 Fall High Human Use vs. Low Human Use 0.593 Winter High Human Use vs. Low Human Use 0.124

The second two-way ANOVA analysis investigated coyote occurrence in relation to life cycle stage and trail type. The within comparison revealed that coyote occurrence was significantly related to life cycle stage, F(1, 56) = 24.982, p <0.001. There were more coyotes on trails during the dispersal season compared to the weaning season (Figure 4.3). Trail type was not significant, F(1, 56) = 0.186, p = 0.668. Thus, there was no difference between coyote occurrence on high and low human use trails.

7 * 6 5 4 3 2 1 0 Average Coyote Abundance/Week -1

Weaning Dispersal

Figure 4.3 Comparison of harmonized mean coyote occurrence per week between the weaning period and the dispersal season on all trail types. The asterisk indicates the significance level when p < 0.01. Error bars represent standard error.

The two-way ANOVA comparison between life cycle stage and trail type was not significant, F(1, 56) = 0.107, p = 0.745. Simple effect testing revealed no significance between the pairwise comparisons (Table 4.5).

Table 4.5 Comparison of coyote occurrence between trail types within the weaning and dispersal life cycle stages.

Life Cycle Stage Trail Type P-value Weaning High Human Use vs. Low Human Use 0.951 Dispersal High Human Use vs. Low Human Use 0.491

The third two-way ANOVA analysis explored coyote occurrence relative to time of day and trail type. The within comparison demonstrated that coyote occurrence was significantly different during periods of the day, F(3, 232) = 21.170, p = < 0.001, but not between trail type, F(1, 232) = 1.116, p = 0.292. Simple effect testing revealed that occurrence was significantly greater during the day and night compared to the morning

and evening twilight periods (Figure 4.4). There was no statistically significant difference in average occurrence between morning and evening twilight as well as between night and day.

2.5 * * 2 * *

1.5 Morning Twilight 1 Daytime Evening Twilight 0.5 Nighttime 0 Average Coyote Abundance/Week -0.5 Morning/ Morning/ Evening/ Evening/ Morning/ Daytime/ Daytime Nighttime Daytime Nighttime Evening Nighttime

Figure 4.4 Comparison of harmonized mean coyote occurrence per week between morning twilight (morning), daytime, evening twilight (evening) and nighttime periods on all trail types. Asterisks indicate significance levels when p < 0.05. Error bars represent standard error.

The comparison of coyote occurrence between trail type and time of day yielded non-significant results, F(3, 232) = 1.974, p = 0.119. However, simple effect testing demonstrated varying levels of significance between pairwise groups (Table 4.6). During the morning twilight, daytime and evening twilight periods coyote occurrence between trail types was not significant. In contrast, during the nighttime period, occurrence was statistically different between high and low human use trails (p = 0.013) (Figure 4.5).

Table 4.6 Comparison of coyote occurrence between trail types within time periods of the day. Time Period Trail Type P-value Morning Twilight High Human Use vs. Low Human Use 0.706 Daytime High Human Use vs. Low Human Use 0.407 Evening Twilight High Human Use vs. Low Human Use 0.940 Nighttime High Human Use vs. Low Human Use 0.013

3.5

* 3

2.5

1.5

0.5 Average Coyote Abundance/Week 0

High Human Use Trails Low Human Use Trails

Figure 4.5 Comparison of harmonized mean coyote occurrence per week between high and low human use trails during the nighttime. The asterisk indicates a significance level when p < 0.05. Error bars represent standard error.

Average coyote occurrence on trails was always greater on high human use trails versus low human use trails (Table 4.7). However, this was never significant except for the trail type comparison during the night. The only period when occurrence was greater on low human use trails compared to high human use trails was during the daytime, though this was not statistically significant.

Table 4.7 Average number of coyotes/week on high and low human use trails during different timeframes. The asterisk indicates significance when p < 0.05. Test Harmonized Mean Harmonized Mean P-Value of Coyote of Coyote Occurrence/Week Occurrence/Week on High Human on Low Human Use Trails Use Trails Summer 1.250 (SE = 1.052) 1.167 (SE = 1.052) 0.956 Fall 6.846 (SE = 1.010) 6.077 (SE = 1.010) 0.593 Winter 10.400 (SE = 1.629) 6.800 (SE = 1.629) 0.124 Weaning 0.889 (SE = 1.263) 0.778 (SE = 1.263) 0.951 Dispersal 6.571 (SE = 0.827) 5.762 (SE = 0.827) 0.491 Morning Twilight 0.467 (SE = 0.312) 0.300 (SE = 0.312) 0.706 Daytime 1.633 (SE = 0.312) 2.000 (SE = 0.312) 0.407 Evening Twilight 0.167 (SE = 0.312) 0.133 (SE = 0.312) 0.940 Nighttime 2.733 (SE = 0.312) 1.633 (SE = 0.312) 0.013*

4.5.2 Research Objective 2: How do human-related disturbances, past coyote trail occurrence, prey, moon phase, season, life cycle stage and time of day relate to coyote occurrence on trails?

Models 1 to 9 were the first models compared using AIC (section 4.4.1.2, p. 43). The top 5 models ranking from highest to lowest were Model 1 (Global), Model 2 (Natural Factors), Model 4 (Time Related), Model 5 (Animals) and Model 3 (Human Related Disturbances). After the first 9 models were created and compared, models 10 to 14 were made. Final results comparing all 14 models are depicted in Table 4.8.

Table 4.8 Comparison of models 1 to 14. A) Model ranks and AIC values Model K -2loglikelihood AIC AICc Δi Weight Ratio Rank 1 (Global) 11 1053.546 1075.546 1075.637 0 0.49 59.04 1 12 (Natural 9 1065.731 1083.731 1083.793 8.156 0.0083 2 Factors + Cyclists + Vehicles) 10 (Natural 8 1082.466 1098.466 1098.516 22.878 5.25*10- 3 Factors + 6 Cyclists) 13 (Natural 8 1109.402 1125.402 1125.452 49.814 7.43*10- 4 Factors + 12 People) 11 (Natural 8 1111.712 1127.712 1127.762 52.124 2.34*10- 5 Factors + 12 Vehicles) 14 (Natural 8 1123.034 1139.034 1139.084 63.446 8.15*10- 6 Factors + 15 Dogs) 2 (Natural 7 1136.286 1150.286 1150.325 74.687 2.98*10- 7 Factors) 17 4 (Time- 4 1244.766 1252.766 1252.78 177.142 1.68*10- 8 Related) 39 5 (Animals) 4 1340.143 1348.143 1348.157 272.519 3.27*10- 9 60 3 (Human- 5 1357.833 1367.833 1367.854 292.216 1.73*10- 10 Related 64 Disturbances) 6 (Coyote 3 1409.983 1415.983 1415.991 340.354 6.09*10- 11 Social 75 Aspects) 7 (People 3 1430.957 1436.957 1436.965 361.328 1.69*10- 12 and Dogs) 79 8 (People) 2 1438.449 1442.449 1442.453 366.816 1.08*10- 13 80 9 (Moon 2 1495.829 1499.829 1499.833 424.196 3.76*10- 14 Phase) 93 Note. Sample size for all models was 2898.

B) Beta coefficients of ratio variables Model # Cyclists # People # Dogs # Prey # Vehicles

1 -1.502* -0.019 -0.032 -0.983* -0.678* 12 -1.559* NA NA -0.924* -0.629* 10 -1.605* NA NA -0.871* NA 13 NA -0.036* NA -0.863* NA 11 NA NA NA -0.836* -0.705* 14 NA NA -0.132* -0.827* NA 2 NA NA NA -0.781* NA 4 NA NA NA NA NA 5 NA NA -0.226* -0.540* NA 3 -1.691* -0.027 0.059 NA -0.621* 6 NA NA NA NA NA 7 NA -0.074* 0.172* NA NA 8 NA -0.049* NA NA NA 9 NA NA NA NA NA Note. Significant beta coefficients (p < 0.05) are starred.

54 C) Beta coefficients of categorical variables

Model Season Life Cycle Time of Day Coyote Moon Phase Stage Occurrence in Past Day Summer Fall Weaning Morning Daytime Evening Absence New Crescent Quarter Gibbous Twilight Twilight Moon Moon Moon Moon 1 -0.452 -0.448* -1.210* -0.795* -1.450* -2.662* -1.262* 0.028 0.401 0.262 0.473

12 -0.524 -0.451* -1.183* -0.794* -1.680* -2.650* -1.179* -0.33 -.391 0.239 0.485

10 -0.567 -0.487* -1.249* -0.841* -1.838* -2.620* -1.218* 0.048 0.478 0.313 0.556

13 -0.686 -0.622* -1.390* -0.807* -1.681* -2.708* -1.330* 0.095 0.461 0.340 0.572

11 -0.894* -0.630* -1.110* -0.740* -1.849* -2.743* -1.275* -0.059 0.345 0.255 0.482 55

14 -0.882* -0.684* -1.375* -0.796* -1.838* -2.693* -1.374* 0.047 0.444 0.331 0.554

2 -0.991* -0.697* -1.295* -0.798* -2.069* -2.688* -1.320* 0.012 0.438 0.338 0.560

4 -1.497* -0.781* -1.117* -0.932* -1.787* -2.239* NA NA NA NA NA 5 NA NA NA NA NA NA -1.840* NA NA NA NA 3 NA NA NA NA NA NA NA NA NA NA NA 6 NA NA NA NA NA NA -1.750* 0.164 0.360 0.190 0.383 7 NA NA NA NA NA NA NA NA NA NA NA 8 NA NA NA NA NA NA NA NA NA NA NA 9 NA NA NA NA NA NA NA 0.274 0.491 0.236 0.556 Note. There are no coefficients for reference categories. Significant beta coefficients (p < 0.05) are starred.

The Global model predicted coyote occurrence better than the other 13 models. Its logistic regression equation (Equation 2) demonstrated that the probability of coyotes decreased as the number of cyclists increased (Wald x2 (1) = 9.063, p = 0.003). This pattern was also seen with the number of people (Wald x2 (1) = 1.664, p = 0.197), the number of dogs (Wald x2 (1) = 0.176, p = 0.675), the number of prey (Wald x2 (1) = 55.055, p < 0.001) and the number of vehicles (Wald x2 (1) = 9.754, p = 0.002).

ln[p/(1-p)] = 0.876 – 1.502(Number of Cyclists) – 0.019(Number of People) – 0.032(Number of Dogs) – 0.983(Number of Prey) – 0.678(Number of Vehicles) – 0.452(Summer) – 0.448(Fall) – 1.210(Weaning) – 0.795(Morning Twilight) – 1.450(Daytime) – 2.662(Evening Twilight) – 1.262(Coyotes Absent in Past 24 Hours) + 0.028(New Moon) + 0.401(Crescent Moon) + 0.262(Quarter Moon) + 0.473(Gibbous Moon) (2)

When examining effects of categorical variables in the Global model, groups within each variable were compared to a reference category. Reference categories for Season, Life Cycle Stage, Time of Day, Coyote Occurrence in the Past Day and Moon Phase were winter, the dispersal season, nighttime, coyote presence within the last day and full moon respectively. Thus, according to Equation 2 coyote occurrence was less in the summer (Wald x2 (1) = 1.465, p = 0.226) and fall (Wald x2 (1) = 5.336, p = 0.021) compared to the winter. Occurrence was also less during the weaning season compared to the dispersal season (Wald x2 (1) = 7.472, p = 0.006). Additionally, probability of occurrence increased when coyotes were on trails within the last 24 hours (Wald x2 (1) = 37.969, p < 0.001). Regarding the time of day, the likelihood of coyote occurrence on trails was greater during the nighttime compared to the daytime (Wald x2 (1) = 53.951, p < 0.001), morning twilight (Wald x2 (1) = 7.151, p = 0.007) and evening twilight periods (Wald x2 (1) = 49.168, p < 0.001). Lastly, coyotes were more likely to be on trails during the new moon (Wald x2 (1) = 0.005, p = 0.943), crescent moon (Wald x2 (1) = 1.457, p = 0.227), quarter moon (Wald x2 (1) = 0.556, p = 0.456) and gibbous moon (Wald x2 (1) = 2.032, p = 0.154) compared to the full moon.

4.6 Discussion

4.6.1 Research Objective 1: Is coyote occurrence less on high human use trails compared to low human use trails? Does coyote occurrence on high versus low human use trails fluctuate across different seasons, life cycle stages and times of the day?

There was no significant difference between coyote occurrence on high and low human use trails. This was surprising, as I had assumed that coyote occurrence would be less on high human use trails based on past studies, which demonstrated a decrease in coyote occurrence relative to high levels of human activity (George & Crooks, 2006; Porter, 2013; Riley et al., 2003). Although my results do not explain causality, one possibility for coyote occurrence on high human use trails in GRPP could be because these trails provided coyotes with quick access to quality resources. In this case, proximity to people may have been worth the attainment of prey, prime denning habitat, etc. Season was significantly related to coyote occurrence, regardless of trail type. Specifically, more coyotes occurred on trails in the fall and winter compared to the summer. One possible explanation for coyote increase on trails could be related to a decrease in the number of park visitors as winter neared, which was observed during fieldwork and photo analysis. Another explanation may be related to the coyote’s life cycle. Coyote occurrence was significantly greater on trail systems during the dispersal season compared to the weaning season. Because the dispersal season occurs over the fall and winter (Lukasik & Alexander, 2012; Tesky, 1995), coyotes may have passed camera stations more frequently as they dispersed throughout the park. Coyote increase on trails may have also been related to snowfall, which occurred in the fall and winter. During fieldwork, I found that snow on high human use trails was heavily packed down, which allowed for easier travel compared to walking through deep snow. There were also areas on low human use trails where wildlife trail use had helped reduce the amount of snow cover, though snow on low human use trails was never as packed down as snow on high human use trails. Consequently, it is possible that trails provided coyotes with a more efficient means of travel during the winter. Wolves have demonstrated similar trail preferences, traveling on snow-covered trails that were packed down by people to allow

for easier movement (Paquet, Alexander, & Donelon, 2010). Coyotes may have also occurred on trails more during fall and winter as a result of food availability. While performing fieldwork, I observed a high abundance of Richardson’s ground squirrels during the summer. Coyotes likely predated this rodent since small mammals make up the largest component of their diet (Fortin-McCuaig, 2012; Lukasik & Alexander, 2012). However, in the winter, Richardson’s ground squirrels hibernate (Michener, 1983). This may have made it increasingly difficult to find prey as temperatures decreased. Should high human use trails allow for easier travel in the fall and winter, coyotes may have used trail systems to more efficiently obtain other food sources during these seasons. Coyote occurrence was significantly greater on trails during the day and night compared to the morning and evening twilight periods, regardless of trail type. Greater coyote trail occurrence during the night was not surprising because coyotes may become more nocturnal in urban areas, possibly to avoid people (Fox & Papouchis, 2005; Gehrt et al., 2009). It was unexpected that coyote trail occurrence was greater during the daytime compared to the twilight periods in GRPP. Upon further analysis, I discovered that average coyote occurrence on low human use trails was greater than occurrence on high human use trails during the daytime, though this was not statistically significant. Thus, it is possible that coyotes adapt to humans by increasing nocturnal activity as well as shifting their spatial occurrence. There was no statistically significant difference between coyote occurrence on high and low human use trails during the daytime, morning twilight and evening twilight periods. However, there were significantly more coyotes on high human use trails during the nighttime than on low human use trails. Coyotes may have taken advantage of high human use trails at night when there were less park visitors. As described previously, snow on high human use trails was more packed down compared to low human use trails, allowing for more efficient movement. When snow was absent, I also found high human use trails easier to travel on because of their flat, wide characteristics. In contrast, low human use trails were often uneven and narrow, which made travel difficult. Thus, there is the potential that coyotes used high human use trails more than low human use trails to increase their efficiency of movement during the nighttime.

4.6.2 Research Objective 2: How do human-related disturbances, past coyote trail occurrence, prey, moon phase, season, life cycle stage and time of day relate to coyote occurrence on trails?

The top ranked model was the Global model, which explained 59.04 percent more than the second best model. The Global model demonstrated that as the number of cyclists (Wald x2 (1) = 9.063, p = 0.003) increased on trails, probability of coyote occurrence decreased. This relationship also occurred between coyotes and cyclists during the summer and fall in Cape Breton Highlands National Park of Canada (CBHNPC), Nova Scotia (Porter, 2013). Coyote occurrence also decreased relative to the number of people (Wald x2 (1) = 1.664, p = 0.197) and domestic dogs (Wald x2 (1) = 0.176, p = 0.675), though these results were not statistically significant. Despite their statistical non-significance, the inclusion of these independent variables in the top model potentially demonstrates that the number of people and dogs may be related to coyote occurrence, even if only at a small level. Relationships between coyotes and dogs/people have been highlighted in other studies. For instance, coyote activity significantly decreased relative to human activity on the Nature Reserve of Orange County (NROC), California (George & Crooks, 2006). Furthermore, in CBHNPC, a significant negative relationship existed between coyote and hiker activity during the fall and summer, while a significant negative correlation occurred between coyote and domestic dog activity in the winter (Porter, 2013). Should coyote occurrence be related to dogs/people in GRPP, I believe that this relationship may be dependent on the presence of other independent variables. When the Number of Dogs and the Number of People variables were modelled alone (Model 7: People & Dogs, Model 8: People), they were poor predictors of coyote occurrence. Yet, they were included in the top model when other human-related variables (cyclists and vehicles) were present. It is possible that the presence of cyclists and vehicles may be needed for a relationship to exist between coyotes and dogs/people. The results of Objectives 1 and 2 seem to contradict one another. There was no significant difference between coyote occurrence on high and low human use trails in Objective 1. However, the Global model in Objective 2 demonstrated that there could potentially be a negative relationship between coyote occurrence and people. If this is the 59

case, I would have assumed that coyote occurrence would be greater on low human use trails compared to high human use trails. Combining the results of both Objectives, I believe that coyotes may have used high human use trails when people were absent, which was observed during photo analysis. Thus, coyotes may have temporally avoided people, but at a finer scale than what was measured in this study. If this assumption is correct, then it is possible that coyotes were more vigilant near high human use trails to avoid people. Although I did not study coyote behavior, fieldwork observations may suggest an increased level of coyote vigilance in areas of higher human activity. For instance, while walking high human use trails during fieldwork, I spotted a total of 2-3 coyotes, which were several hundred metres from the trail. In contrast, coyotes on low human use trails were much closer to me and were spotted more frequently. They also seemed “surprised” to see a human; once they noticed me, they tended to quickly run off. If coyote vigilance increases in areas where human activity is high, then this may imply that the benefits associated with using high human use trails, such as access to food resources, outweighs the costs of being near people. This has been demonstrated by wolves (Paquet, Wierzchowski, & Callaghan, 1996). Although wolves preferred spaces that lacked people, they were still observed using human-populated spaces if habitat quality was high. Thus, there is the possibility that an increase in coyote awareness near high human use trails allows coyotes to better access resources that are located closer to people, though future research is needed to explore this further. Coyote occurrence was also negatively related to the number of vehicles on trails (Wald x2 (1) = 9.754, p = 0.002). In contrast, there was no relationship between coyote and vehicle activity in NROC, California (George & Crooks, 2006). A potential explanation could be that coyotes in GRPP are less habituated to vehicles than coyotes in NROC. There is also the possibility that coyotes in GRPP have had more negative interactions with vehicles compared to coyotes in NROC. Driving along highway 1A, located just north of GRPP, it is not uncommon to see coyote road kill. As well, during fieldwork, I spotted a coyote that seemed to be paralyzed from the waist down. There is the potential that the coyote was struck by a car. Thus, coyotes in GRPP may have learned to avoid vehicles. The negative relationship between coyote and vehicle

occurrence in GRPP may have also existed because vehicles were not associated with a “reward”. While working a summer job for Alberta Parks in the Kananaskis region, I learned that park visitors occasionally feed wildlife from their vehicles. As a result, wildlife is sometimes attracted to cars/trucks in search of food. However, in GRPP, park visitors are not allowed vehicular access to trail systems; most vehicles on trails are driven by Alberta Parks staff. Because coyotes are likely not fed from vehicles in GRPP, vehicles may not be an attractant. This, in combination with possible coyote-related vehicle injury/mortality, may potentially explain why there is a negative relationship between coyote and vehicle occurrence in GRPP. Future research could explore the causality behind this negative relationship as well as why differences between coyote- vehicle occurrence patterns differ between NROC and GRPP. Interestingly, coyote occurrence decreased on trails in association with prey (Wald x2 (1) = 55.055, p < 0.001). This was unexpected, as I believed coyotes would be near food sources, assuming prey occurred on trails. However, other research has demonstrated that co-occurrence between carnivore and prey species is minimal, possibly due to prey avoidance of predators (Muhly, Semeniuk, Massolo, Hickman, & Musiani, 2011). One way avoidance is achieved is through increased prey vigilance when carnivores are near, as demonstrated by elk (Wolff & Van Horn, 2003). Cow elk were more attentive in areas where carnivores were present compared to cow elk that inhabited areas that lacked predators. Likewise, elk in habitats that contained predators spent less time in open areas. Prey may also gravitate to areas that predators avoid (Muhly et al., 2011). For example, humans tend to displace carnivores more than prey species, which provides safer spaces for prey. Thus, in GRPP there is the potential that prey selected for sites that offered protection from coyotes, which could explain why there was a negative beta-coefficient for prey in the top logistic regression model. Objective 2 demonstrated that coyote occurrence on trails was less in the summer (Wald x2 (1) = 1.465, p = 0.226) and fall (Wald x2 (1) = 5.336, p = 0.021) compared to the winter. As described earlier, this may potentially have been related to a decrease in human trail use as winter neared. Similar to Objective 1, coyote occurrence was greater during the dispersal season compared to the weaning season (Wald x2 (1) = 7.472, p =

0.006); trails may have provided easy means of movement during the dispersal season. Additionally, probability of coyote occurrence was significantly greater during the night compared to during the day (Wald x2 (1) = 53.951, p < 0.001), morning twilight (Wald x2 (1) = 7.151, p = 0.007) and evening twilight periods (Wald x2 (1) = 49.168, p < 0.001). Coyote activity may increase on trails during the night to avoid people, as demonstrated by other research (Fox & Papouchis, 2005; Gehrt et al., 2009). Coyote occurrence significantly increased on trails when coyote(s) were present on trails within the past 24 hours (Wald x2 (1) = 37.969, p < 0.001). While it is possible to identify individual coyotes in images, I was not able to do this due to time constraints and the variability in camera photo quality. Hence, coyotes in subsequent photos may have been conspecifics or the same coyote(s). Because coyotes may scent mark their territory quickly after interactions with other coyotes (Bowen & Cowen, 1980), this may have explained why coyote trail occurrence increased when coyotes used trails within the past 24 hours. Another possibility is that coyotes were finding or courting a mate; during the breeding season coyotes of the opposite sex tend to travel closer together (Andelt, 1985). There could also have been a kill left in the area or carrion, which continuously attracted coyotes (Hein & Andelt, 1996). Moon phases and coyote occurrence were not significantly related. However, the probability of coyote trail occurrence was greater during the new moon (Wald x2 (1) = 0.005, p = 0.943), crescent moon (Wald x2 (1) = 1.457, p = 0.227), quarter moon (Wald x2 (1) = 0.556, p = 0.456) and gibbous moon (Wald x2 (1) = 2.032, p = 0.154) compared to the full moon. Other research that has explored coyote association with moon phase demonstrated that “group howling” occurs less frequently when there is an increase in moonlight (Bender, Bayne, & Brigham, 1996). It was postulated that this may be because of prey. Darker nights could prevent coyotes from finding small mammals as effectively as on moonlit nights. As a result, coyotes may be more likely to trespass on another coyote’s territory to find food. Thus, howling may have been used as a means of territorial defense (Gese, 2004) as well as potentially locating other conspecifics (Mitchell, Makagon, Jaeger, & Barrette, 2006). If this assumption is correct, then it could explain why coyote occurrence increased on trails during darker nights; coyotes may

have used trails to increase the efficiency of finding food when prey was less visible. Because moon phase was included in the top model despite its statistical non- significance, it is possible that a meaningful association between coyote trail occurrence and moon phase exists. Future research could further explore potential relationships between these two variables. When comparing the models, there seems to be a threshold regarding human disturbances. The top 7 models all included the natural factor variables (Season, Life Cycle Stage, Time of Day, Coyote Occurrence in Past Day, Moon Phase and Number of Prey) with the addition of various human disturbance variables. It seems that cyclists predicted coyote occurrence better than other human-related disturbances, as they were in the top three models. The relationship between coyote occurrence and vehicles was interesting. Vehicles and natural factors (Model 11) did not predict coyote occurrence as well as people and natural factors (Model 13). Yet, the model including vehicles, cyclists and natural factors (Model 12) predicted coyote occurrence better than Model 13. This suggests that vehicles alone do not predict coyote occurrence as well as vehicles paired with cyclists. Although natural factors were good predictors of coyote occurrence, they must be associated with human disturbance(s) to create the best fitting models.

4.7 Conclusion

This study demonstrated that coyotes potentially shift their occurrence relative to human occurrence. Not only did coyotes use trails systems more during the winter and fall compared to the summer, but coyotes were also more prevalent on trail systems during the night when there were fewer park visitors. Specifically, coyote occurrence was greater on high human use trails during the nighttime compared to low human use trails. Furthermore, coyotes may have been more vigilant on/near high human use trails to avoid people, though this was not specifically tested. Lastly, coyote trail occurrence decreased when other human-related disturbances (i.e. vehicles and cyclists) were present. Coyote occurrence was greater on trails during the dispersal season compared to the weaning season, which offers another possible explanation regarding higher coyote trail occurrence during the fall and winter. Furthermore, coyote occurrence increased on 63

trails when coyote(s) had used trails within the previous day. Prey occurrence was negatively related to coyote occurrence, possibly because prey actively avoided coyotes. Thus, there were a variety of factors that predicted coyote occurrence on trail systems.

Chapter Five: Management Recommendations and Future Research

5.1 Management Recommendations

This research provided important baseline data for GRPP. Based on results, I have three key recommendations:

1) Cyclists predicted coyote trail occurrence more than other human-related factors; as the number of cyclists increased, the probability of coyote occurrence decreased. In terms of coyote survival needs, it may be beneficial for coyotes if cycling access is reduced during seasonal coyote resource-stress times, such as the denning season.

2) Many dogs observed in photos were off leash. Because the Global model demonstrated a potential negative relationship between coyote and domestic dog occurrence, it may be advantageous for coyotes if dogs are kept leashed.

Recommendations 1) and 2) were described based on benefits to coyotes. However, reducing the number of cyclists and keeping dogs leashed during times when coyotes are providing for pups, may also help prevent human-coyote conflict. The pup-rearing season is the period when most human-coyote conflict occurs (Lukasik & Alexander, 2011). Should coyotes feel threatened by dogs/cyclists during this time, then there may be the potential for a negative encounter. As a result, respecting coyote survival needs may help prevent conflict.

3) Increasing public awareness on how to coexist with coyotes could help prevent human-coyote conflict. People may not know how frequently coyotes use the trail systems. In fact, coyote occurrence was not significantly different between high and low human use trails. Additionally, coyote occurrence increased on trails during the fall, the winter and the dispersal season as well as during the daytime and nighttime. Knowing that coyotes use trails may help people act responsibly

by leashing their dogs and properly disposing of their garbage. As well, it is important for park visitors to understand what to do should they encounter a coyote. Appropriate human behavior may help reduce potential conflict.

This research has revealed negative relationships that exist between coyotes and human-related disturbances. Should these relationships be related to coyote avoidance of people, then it is an encouraging sign that coexistence in GRPP may be very possible.

5.2 Future Research

Unfortunately, not all cameras collected data consistently. As a result, all studies had reduced timeframes of analyses. Future research could explore human-coyote co- occurrence during the other life cycle stages, including the entire length of the pup- rearing season as well as the breeding and birthing seasons. Specifically, because the pup-rearing stage is when most human-coyote conflict occurs (Lukasik & Alexander, 2011), it may be important to study coyote spatial patterns during this time. GRPP is still undergoing trail construction (Glenbow Ranch Provincial Park Foundation, 2015). The goal is to link Cochrane and Calgary with the Calgary to Cochrane (C TO C) Trail. It would be interesting to explore coyote occurrence in areas where paths will be built. Where is coyote occurrence greatest in the park? Is it possible to build trails around these areas? What is coyote occurrence like before and after trails are created? Additionally, understanding how human use on trails affects coyote behavior and life history is important. Although coyotes used high human use trails, does human presence negatively affect them? If it does, what do coyotes gain from using trails? What are the costs of using trails? Does coyote trail occurrence decrease during the weaning season to avoid people or because more time is spent near the den? Should future studies demonstrate that human presence has a detrimental effect on coyotes, then it might be beneficial to implement specific trail closures while coyotes experience seasonal resource-stress. Future research could also extend the methods study. Since scat surveys did not detect coyote occurrence during all spring biweekly periods, it would be interesting to 66

explore why that occurred and if that pattern continues on into the summer. Additionally, lengthening the analysis on both high and low human use trails could help highlight what method is best for each trail type across different seasons.

References

Agresti, A., & Franklin, C. (2007). Statistics: The art and science of learning from data. Upper Saddle River, New Jersey: Pearson Education, Inc. Alberta Agriculture and Forestry. (2009). Overview of the Pest and Nuisance Control Regulation. Retrieved from http://www1.agric.gov.ab.ca/$department/deptdocs.nsf/all/rsb12565 Alberta Environment and Parks. (2014a). Coyote (Canis latrans). Retrieved from http://aep.alberta.ca/fish-wildlife/wild-species/mammals/wild-dogs/coyote.aspx Alberta Environment and Parks. (2014b). Hunting in Alberta. Retrieved from http://aep.alberta.ca/fish-wildlife/fishing-hunting-trapping/hunting- alberta/default.aspx Alberta Parks. (2015). Glenbow Ranch Provincial Park: History. Retrieved from http://www.albertaparks.ca/glenbow-ranch/information-facilities/history.aspx Alberta Sustainable Resource Development. (2005). Mapping planes in Alberta. Retrieved from http://www.servicealberta.gov.ab.ca/pdf/ltmanual/Fact_Sheet_10.pdf Alexander, S. M. (2014). Unpublished raw data. Department of Geography. University of Calgary. Alexander, S. M. (2015). Unpublished raw data. Department of Geography. University of Calgary. Alexander, S. M., & Quinn, M. S. (2011). Coyote (Canis latrans) interactions with humans and pets reported in the Canadian print media (1995–2010). Human Dimensions of Wildlife, 16(5), 345-359. Alexander, S. M., & Quinn, M. S. (2012). Portrayal of interactions between humans and coyotes (Canis latrans): Content analysis of Canadian print media (1998-2010). Cities and the Environment, 4(1), Article 9. Andelt, W. F. (1985). Behavioral ecology of coyotes in south Texas. Wildlife Monographs, 94, 3-45. Baker, R. O. (2007). A review of successful urban coyote management programs implemented to prevent or reduce attacks on humans and pets in southern California. Paper presented at the Wildlife Damage Management Conferences- Proceedings. Baker, R. O., & Timm, R. M. (1998). Management of conflicts between urban coyotes and humans in southern California. Paper presented at the Proceedings of the Eighteenth Vertebrate Pest Conference. Bekoff, M. (1977). Canis latrans. Mammalian Species, (79), 1-9. Bekoff, M., & Gese, E. M. (2003). Coyote (Canis latrans). In G. A. Feldhamer, B. C. Thompson, & J. A. Chapman (Eds.), Wild mammals of North America: Biology, management, and conservation (2 ed., pp. 467-481). Baltimore, Maryland: The Johns Hopkins University Press. Bender, D., J., Bayne, E. M., & Brigham, M. R. (1996). Lunar condition influences coyote (Canis latrans) howling. American Midland Naturalist, 136(2), 413-417.

Berger, K. M. (2006). Carnivore-livestock conflicts: Effects of subsidized predator control and economic correlates on the sheep industry. Conservation Biology, 20(3), 751-761. Berger, K. M., & Gese, E. M. (2007). Does interference competition with wolves limit the distribution and abundance of coyotes? Journal of Animal Ecology, 76(6), 1075-1085. Bowen, W. D. (1978). Social organization of the coyote in relation to prey size. (Doctor of Philosophy Dissertation), University of British Columbia, Vancouver, British Columbia. Retrieved from https://open.library.ubc.ca/cIRcle/collections/ubctheses/831/items/1.0094757 Bowen, W. D. (1981). Variation in coyote social organization: The influence of prey size. Canadian Journal of Zoology, 59(4), 639-652. Bowen, W. D. (1982). Home range and spatial organization of coyotes in Jasper National Park, Alberta. The Journal of Wildlife Management, 46(1), 201-216. Bowen, W. D., & Cowen, I. M. (1980). Scent marking in coyotes. Canadian Journal of Zoology, 58(4), 473-480. Brichieri-Colombi, T. (2012). Use of non-invasive methods to examine species distribution relative to a highway in the Calakmul region, Mexico. (Master of Science Dissertation), University of Calgary, Calgary, Alberta. Burnham, K. P., & Anderson, D. R. (2002). Model selection and multimodel inference: A practical information-theoretic approach (2 ed.). New York, United States: Springer. Burt, W. H. (1943). Territoriality and home range concepts as applied to mammals. Journal of Mammalogy, 24(3), 346-352. Camenzind, F. J. (1978). Behavioral ecology of coyotes (Canis latrans) on the National Elk Refuge Jackson, Wyoming. In M. Bekoff (Ed.), Coyotes: Biology, behavior, and management (pp. 267-294). New York, New York: Academic Press. Canada Safety Council. (2005). Aggressive dogs threaten public safety. Retrieved from https://canadasafetycouncil.org/child-safety/aggressive-dogs-threaten-public- safety Cattet, M., Boulanger, J., Stenhouse, G., Powell, R. A., & Reynolds-Hogland, M. J. (2008). An evaluation of long-term capture effects in ursids: Implications for wildlife welfare and research. Journal of Mammalogy, 89(4), 973-990. CBC News. (2014). Coyote problem declining in parts of Nova Scotia. Retrieved from http://www.cbc.ca/1.2749868 CBC News. (2015). Coyote hunt with cash prizes draws controversy, threats in Alberta. Retrieved from http://www.cbc.ca/news/canada/edmonton/coyote-hunt-with-cash- prizes-draws-controversy-threats-in-alberta-1.2894093 Clegg, S. M., & Owens, I. P. F. (2002). The "island rule" in birds: Medium body size and its ecological explanation. Proceedings of the Royal Society of London B. Cook, S. M., Awmack, C. S., Murray, D. A., & Williams, I. H. (2003). Are honey bees' foraging preferences affected by pollen amino acid composition? Ecological Entymology, 28(5), 622-627.

Crabtree, R. L., & Sheldon, J. W. (1999a). Coyotes and canid coexistence in Yellowstone. In T. W. Clark, A. P. Curlee, S. C. Minta, & P. Kareiva, IV (Eds.), Carnivores in ecosystems: The Yellowstone experience (pp. 127-163). New Haven, Connecticut: Yale University Press. Crabtree, R. L., & Sheldon, J. W. (1999b). The ecological role of coyotes on Yellowstone's northern range. Yellowstone Science, 7(2), 15-23. Crooks, K. R., & Soulé, M. E. (1999). Mesopredator release and avifaunal extinctions in a fragmented system. Nature, 400(6744), 563-566. Crooks, K. R., Suarez, A. V., Bolger, D. T., & Soulé, M. E. (2001). Extinction and colonization of birds on habitat islands. Conservation Biology, 15(1), 159-172. Danner, D. A., & Smith, N. S. (1980). Coyote home range, movement, and relative abundance near a cattle feedyard. The Journal of Wildlife Management, 44(2), 484-487. Darimont, C. T., Carlson, S. M., Kinnison, M. T., Paquet, P. C., Reimchen, T. E., & Wilmers, C. C. (2009). Human predators outpace other agents of trait change in the wild. Proceedings of the National Academy of Sciences, 106(3), 952-954. Data Resource Centre, U. o. G. (2012). Defining projections in ArcMap 10.1. Retrieved from http://www.lib.uoguelph.ca/sites/default/files/defining_projections_arcgis_ten_on e.pdf Dell'Amore, C. (2014). Downtown coyotes: Inside the secret lives of Chicago's predator. Retrieved from http://news.nationalgeographic.com/news/2014/11/141121- coyotes-animals-science-chicago-cities-urban-nation Environment Canada. (2013). Canada's top 10 weather stories for 2011. Retrieved from http://www.ec.gc.ca/meteo-weather/default.asp?lang=En&n=774B5B53-1 Environment Canada. (2015). Canadian climate normals 1981-2010 station data. Retrieved from Final Copy of Thesis_Feb 7, 2016.docx Fall, M. W. (1990). Control of coyote predation on livestock- Progress in research and development. Paper presented at the Proceedings of the Fourteenth Vertebrate Pest Conference. Fedriani, J. M., Fuller, T. K., Sauvajot, R. M., & York, E. C. (2000). Competition and intraguild predation among three sympatric carnivores. Oecologia, 125(2), 258- 270. Field, A. (2009). Discovering statistics using SPSS (3 ed.). London, England: SAGE Publications Ltd. Field, A., Miles, J., & Field, Z. (2012). Discovering statistics using R. London, England: SAGE Publications Ltd. Fortin-McCuaig, M. (2012). Spatial and seasonal differences in the diets of urban and rural coyotes (Canis latrans) in the Calgary, AB vicinity. (Master of Science Thesis), University of Calgary, Calgary, Alberta. Retrieved from http://search.proquest.com.ezproxy.lib.ucalgary.ca/pqdtglobal/docview/10199887 52/2892D6B189EB4C93PQ/7?accountid=9838 Fox, C. H. (2012). Help stop "coyote & wolf derby" in Salmon, Idaho. Retrieved from http://hosted.verticalresponse.com/567808/7d9b2dc9f8/1557503015/2f6d9a2a36/

Fox, C. H. (2014). [Personal communication, Founder & Executive Director, Project Coyote]. Fox, C. H., & Papouchis, C. M. (2005). Coyotes in our midst. Sacramento, California: Animal Protection Institute. Gehrt, S. D., Anchor, C., & White, L. A. (2009). Home range and landscape use of coyotes in a metropolitan landscape: Conflict or coexistence? Journal of Mammalogy, 90(5), 1045-1057. GeoKs. (2012). Glenbow Ranch Provincial Park. Retrieved from https://thegeoks.wordpress.com/2012/06/02/glenbow-ranch-provincial-park/ George, S. L., & Crooks, K. R. (2006). Recreation and large mammal activity in an urban nature reserve. Biological Conservation, 133(1), 107–117. Gese, E. M. (2004). Coyotes in Yellowstone National Park: The influence of dominance on foraging, territoriality, and fitness. In D. W. MacDonald & C. Sillero-Zubiri (Eds.), The biology and conservation of wild canids (pp. 271-283). Oxford, England: Oxford University Press. Gese, E. M., & Grothe, S. (1995). Analysis of coyote predation on deer and elk during winter in Yellowstone National Park, Wyoming. American Midland Naturalist, 133(1), 36-43. Gibeau, M. L. (1993). Use of urban habitats by coyotes in the vicinity of Banff, Alberta. (Master of Science Thesis), University of Montana, Missoula, Montana. Gier, H. T. (1968). Coyotes in Kansas. Retrieved from Manhattan, Kansas: http://www.k- state.edu/historicpublications/Pubs/SB393.pdf Glenbow Ranch Park Foundation. (n.d.-a). Glenbow Ranch Provincial Park plant checklist. Retrieved from http://www.grpf.ca/wp-content/uploads/2014/05/GRPP- Plant-Species-Checklist.pdf Glenbow Ranch Park Foundation. (n.d.-b). Glenbow Ranch Provincial Park Wildlife Checklist. Retrieved from http://www.grpf.ca/wp- content/uploads/2015/05/GRPP-Wildlife-Species-Checklist.pdf Glenbow Ranch Provincial Park Foundation. (2015). Calgary to Cochrane Trail. Retrieved from http://www.grpf.ca/ctoctrail Gompper, M. E. (2002). Top carnivores in the suburbs? Ecological and conservation issues raised by colonization of north-eastern North America by coyotes. BioScience, 52(2), 185-190. Gompper, M. E., Kays, R. W., Ray, J. C., Lapoint, S. D., Bogan, D. A., & Cryan, J. R. (2006). A comparison of noninvasive techniques to survey carnivore communities in northeastern North America. Wildlife Society Bulletin, 34(4), 1142-1151. Gould, J. L., & Gould, G. F. (2002). Biostats basics: A student handbook. New York, United States: W. H. Freeman and Company. Agricultural Pests Act: Pest and Nuisance Control Regulation, (2001). Government of Alberta. (2010). Coyote predation control manual and study guide. Retrieved from http://www1.agric.gov.ab.ca/general/progserv.nsf/all/pgmsrv403/$FILE/manual- study-guide.pdf. Halfpenny, J., & Biesiot, E. (1986). A field guide to mammal tracking in North America (2 ed.). Boulder, Colorado: Johnson Printing Company.

Harrison, D. J., & Gilbert, J. R. (1985). Denning ecology and movements of coyotes in Maine during pup rearing. Journal of Mammalogy, 66(4), 712-719. Hein, E. W., & Andelt, W. F. (1996). Coyote visitations to experimentally-placed deer carrion. The Southwestern Naturalist, 41(1), 48-53. Henke, S. E., & Bryant, F. C. (1999). Effects of coyote removal on the faunal community in western Texas. The Journal of Wildlife Management, 63(4), 1066-1081. Howell, D. C. (2014). Fundamental statistics for the behavioral sciences (8 ed.). Belmont, California: Wadsworth, Cengage Learning. IBM. (2014). Significant interaction in ANOVA: How to obtain a simple effects test. Retrieved from http://www-01.ibm.com/support/docview.wss?uid=swg21475404 IUCN. (2015). The IUCN red list of threatened species. Retrieved from http://www.iucnredlist.org Johnston, H. (2015). [Personal communication, Conservation Officer, Cochrane-Glenbow District]. Kamler, J. F., Ballard, W. B., Gilliland, R. L., & Mote, K. (2003). Spatial relationships between swift foxes and coyotes in northwestern Texas. Canadian Journal of Zoology, 81(1), 168-172. Kamler, J. F., Ballard, W. B., Lemons, P. R., Gilliland, R. L., & Mote, K. (2005). Home range and habitat use of coyotes in an area of native prairie, farmland and CRP fields. The American Midland Naturalist, 153(2), 396-404. Kelly, M. J., & Holub, E. L. (2008). Camera trapping of carnivores: Trap success among camera types and across species, and habitat selection by species, on Salt Pond Mountain, Giles County, Virginia. Northeastern Naturalist, 15(2), 249-262. Kitchen, A. M., Gese, E. M., & Schauster, E. R. (2000). Changes in coyote activity patterns due to reduced exposure to human persecution. Canadian Journal of Zoology, 78(5), 853-857. Knowlton, F. F. (1972). Preliminary interpretations of coyote population mechanics with some management implications. The Journal of Wildlife Management, 36(2), 369- 382. Koehler, G. M., & Hornocker, M. G. (1991). Seasonal resource use among mountain lions, bobcats, and coyotes. Journal of Mammalogy, 72(2), 391-396. Laerd Statistics. (2013a). Binomial logistic regression using SPSS Statistics. Retrieved from https://statistics.laerd.com/spss-tutorials/binomial-logistic-regression-using- spss-statistics.php Laerd Statistics. (2013b). One-way ANOVA (cont...). Retrieved from https://statistics.laerd.com/statistical-guides/one-way-anova-statistical-guide- 2.php Laerd Statistics. (2013c). Two-way ANOVA in SPSS Statistics. Retrieved from https://statistics.laerd.com/spss-tutorials/two-way-anova-using-spss-statistics.php Long, R. A., MacKay, P., Zielinski, W. J., & Ray, J. C. (2008). Noninvasive survey methods for carnivores. Washington, D.C.: Island Press. Lukasik, V. M., & Alexander, S. M. (2011). Human-coyote interactions in Calgary, Alberta. Human Dimensions of Wildlife, 16(2), 114-127.

Lukasik, V. M., & Alexander, S. M. (2012). Spatial and temporal variation of coyote (Canis latrans) diet in Calgary, Alberta. Cities and the Environment, 4(1), Article 8. MacDonald, B., D. (2012). Spatial and temporal patterns of habitat use by East Pacific green turtles, Chelonia mydas, in a highly urbanized foraging ground. (Master of Science Thesis), San Diego State University, San Diego, California. Retrieved from http://www.seaturtle.org/PDF/Ocr/MacDonaldBD_2012_MSc.pdf Major, J. T., & Sherburne, J. A. (1987). Interspecific relationships of coyotes, bobcats, and red foxes in western Maine. The Journal of Wildlife Management, 51(3), 606- 616. Martin, C. (2012). A new wildlife portfolio: Coyotes! Retrieved from http://christophermartinphotography.com/category/wildlife/coyotes/ Mccoy, J. C., Ditchkoff, S. S., & Steury, T. D. (2011). Bias associated with baited camera sites for assessing population characteristics of deer. Journal of Wildlife Management, 75(2), 472-477. McDonald, J. H. (2014). Handbook of biological statistics. Retrieved from http://www.biostathandbook.com/ McManus, J. S., Dickman, A. J., Gaynor, D., Smuts, B. H., & Macdonald, D. W. (2015). Dead or alive? Comparing costs and benefits of lethal and non-lethal human- wildlife conflict mitigation on livestock farms. Oryx, 49(4), 687-695. Merkle, J. A., Stahler, D. R., & Smith, D. W. (2009). Interference competition between gray wolves and coyotes in Yellowstone National Park. Canadian Journal of Zoology, 87(1), 56-63. Messier, F., & Barrette, C. (1982). The social system of the coyote (Canis latrans) in a forested habitat. Canadian Journal of Zoology, 60(7), 1743-1753. Michener, G. R. (1983). Spring emergence schedules and vernal behavior of Richardson's ground squirrels: Why do males emerge from hibernation before females? Behavioral Ecology and Sociobiology, 14(1), 29-38. Milner, J. M., Nilsen, E. B., & Andreassen, H. P. (2007). Demographic side effects of selective hunting in ungulates and carnivores. Conservation Biology, 21(1), 36- 47. Mitchell, B. R., Makagon, M. M., Jaeger, M. M., & Barrette, R. H. (2006). Information content of coyote barks and howls. Bioacoustics, 15(3), 289-314. Moore, G. C., & Parker, G. R. (1992). Colonization by the eastern coyote (Canis latrans). In A. H. Boer (Ed.), Ecology and management of the eastern coyote (pp. 23-37). Fredericton, New Brunswick: Wildlife Research Unit. Morey, P. S., Gese, E. M., & Gehrt, S. (2007). Spatial and temporal variation in the diet of coyotes in the Chicago Metropolitan Area. The American Midland Naturalist, 158(1), 147-161. Muhly, T. B., Semeniuk, C., Massolo, A., Hickman, L., & Musiani, M. (2011). Human activity helps prey win the predator-prey space race. PLoS One, 6(3), e17050. National Agricultural Statistics Service. (2005). Sheep and goats death loss. Retrieved from Washington, D.C.: http://usda.mannlib.cornell.edu/usda/nass/sgdl//2000s/2005/sgdl-05-06-2005.pdf

National Agricultural Statistics Service. (2006). Cattle death loss. Retrieved from Washington, D.C.: http://usda.mannlib.cornell.edu/usda/nass/CattDeath//2000s/2006/CattDeath-05- 05-2006.pdf National Geographic Wild. Killed by coyotes. Retrieved from http://natgeotv.com/uk/killed-by-coyotes/about National Research Council Canada. (2015). Sunrise/sunset calculator. Retrieved from http://www.nrc-cnrc.gc.ca/eng/services/sunrise/index.html Nau, R. (2015). Stepwise and all-possible-regressions. Retrieved from http://people.duke.edu/~rnau/regstep.htm Niedzialkowska, M., Jedrzejewski, W., Myslajek, R. W., Nowak, S., Jedrzejewska, B., & Schmidt, K. (2006). Environmental correlates of Eurasian lynx occurrence in Poland- Large scale census and GIS mapping. Biological Conservation, 133(1), 63-69. Nellis, C. H., & Keith, L. B. (1976). Population dynamics of coyotes in central Alberta, 1964-68. The Journal of Wildlife Management, 40(3), 389-399. Noble, G. K. (1939). The role of dominance in the social life of birds. The Auk, 56(3), 263-273. O'Neill, K. (2009). Sask. puts a bounty on the wily coyote. Retrieved from http://www.theglobeandmail.com/news/national/sask-puts-a-bounty-on-the-wily- coyote/article1205136/ Olsen, G. H., Linhart, S. B., Holmes, R. A., Dasch, G. J., & Male, C. B. (1986). Injuries to coyotes caught in padded and unpadded steel foothold traps. Wildlife Society Bulletin, 14(3), 219-223. Onderka, D. K., Skinner, D. L., & Todd, A. W. (1990). Injuries to coyotes and other species caused by four models of footholding devices. Wildlife Society Bulletin, 18(2), 175-182. Ordeñana, M. A., Crooks, K. R., Boydston, E. E., Fisher, R. N., Lyren, L. M., Siudyla, S., . . . Van Vuren, D. H. (2010). Effects of urbanization on carnivore species distribution and richness. Journal of Mammalogy, 91(6), 1322-1331. Ozoga, J. J., & Harger, E. M. (1966). Winter activities and feeding habits of northern Michigan coyotes. The Journal of Wildlife Management, 30(4), 809-818. Paquet, P. C., Alexander, S. M., & Donelon, S. (2010). Influence of anthropogenically modified snow conditions on movements and predatory behaviour of gray wolves. In M. Musiani, L. Boitani, & P. C. Paquet (Eds.), The world of wolves: New perspectives on ecology, behaviour, and management (pp. 157-173). Calgary, Alberta: University of Calgary Press. Paquet, P. C., Wierzchowski, J., & Callaghan, C. (1996). Effects of human activity on gray wolves in the Bow River Valley, Banff National Park, Alberta. In J. Green, C. Pacas, L. Cornwell, & S. Bayley (Eds.), Ecological outlooks project: A cumulative effects assessment and futures outlook of the Banff Bow Valley. Ottawa, Ontario: Department of Canadian Heritage. Person, D. K., & Hirth, D. H. (1991). Home range and habitat use of coyotes in a farm region of Vermont. The Journal of Wildlife Management, 55(3), 433-441.

Petchey, O. L., McPhearson, P. T., Casey, T. M., & Morin, P. J. (1999). Environmental warming alters food-web structure and ecosystem function. Nature, 402(6757), 69-72. Peterson, A. T., & Robins, C. R. (2003). Using ecological-niche modeling to predict Barred Owl invasions with implications for Spotted Owl conservation. Conservation Biology, 17(4), 1161-1165. Platt, M. (2016). Optics, rather than intent, casts pall over annual northwestern Alberta coyote hunting contest. Retrieved from http://www.calgarysun.com/2016/01/04/optics-rather-than-intent-casts-pall-over- annual-northwestern-alberta-coyote-hunting-contest Porter, K. (2013). Spatial overlap between human and coyote (Canis latrans) activity in Cape Breton Highlands National Park of Canada. (Master of Environmental Science Thesis), Dalhousie University, Halifax, Nova Scotia. Retrieved from http://dalspace.library.dal.ca/bitstream/handle/10222/21675/Porter-Kate-MES- RESOURCE-AND-ENVIRONMENTAL-STUDIES-March- 2013.pdf?sequence=1 Reconyx. (2006). Silent Image user Guide. Reconyx. (2013). Hyperfire high performance cameras instruction manual. Reed, S. E. (2011). Non-invasive methods to assess co-occurrence of mammalian carnivores. The Southwestern Naturalist, 56(2), 231-240. ReStore. (2011). Using statistical regression methods in education research. Retrieved from http://www.restore.ac.uk/srme/www/fac/soc/wie/research- new/srme/modules/index.html Riley, S. P. D., Sauvajot, R. M., Fuller, T. K., York, E. C., Kamradt, D. A., Bromley, C., & Wayne, R. K. (2003). Effects of urbanization and habitat fragmentation on bobcats and coyotes in southern California. Conservation Biology, 17(2), 566- 576. Rose, S., Spinks, N., & Canhoto, A. I. (2015). Tests for the assumption that a variable is normally distributed Management research: Applying the principles. Retrieved from http://documents.routledge- interactive.s3.amazonaws.com/9780415628129/Chapter 13 - Tests for the assumption that a variable is normally distributed final_edited.pdf Sarstedt, M., & Mooi, E. (2014). A concise guide to market research: The process, data, and methods using IBM SPSS statistics: Springer-Verlag Berlin Heidelberg. Siegel, S. (1956). Nonparametric statistics for the behavioral sciences. New York: McGraw-Hill Book Company. Sillero-Zubiri, C., Hoffmann, M., & Macdonald, D. W. (2004). Canids: Foxes, wolves, jackals and dogs. Gland, Switzerland and Cambridge, UK: IUCN/SSC Canid Specialist Group. Silva, J. M. C., Uhl, C., & Murray, G. (1996). Plant succession, landscape management, and the ecology of frugivorous birds in abandoned Amazonian pastures. Conservation Biology, 10(2), 491-503.

Silver, S. (2004). Assessing jaguar abundance using remotely triggered cameras. Retrieved from http://www.academia.edu/5141653/Assessing_jaguar_abundance_using_remotely _triggered_cameras_Table_of_Contents Simkiss, D., Edmond, K., Bose, A., Troy, S., & Bassat, Q. (2015). Research methods II: Multivariate analysis: Regression diagnostics. Journal of Tropical Pediatrics. Retrieved from http://www.oxfordjournals.org/our_journals/tropej/online/ma.html Smith, D. W., Stahler, D. R., Guernsey, D. S., Metz, M., Albers, E., Williamson, L., . . . McIntyre, R. (2007). Yellowstone wolf project: Annual report. Retrieved from Yellowstone National Park, Wyoming: http://www.nps.gov/yell/learn/nature/upload/wolfarpt2007.pdf Smith, J. R. (2009). Spatial organization, habitat preference, and management of northern flying squirrels, Glaucoyms sabrinus, in the northern Sierra Nevada. (Master of Science Thesis), University of California, California. Retrieved from http://www.sierraforestlegacy.org/Resources/Conservation/SierraNevadaWildlife/ NorthernFlyingSquirrel/NFS-Smith_2009_Thesis.pdf Statistics Canada. (2010). Population and dwelling counts, for Canada and census subdivisions (municipalities), 2006 and 2001 censuses: 100% data. Retrieved from https://www12.statcan.gc.ca/census-recensement/2006/dp-pd/hlt/97- 550/Index.cfm?TPL=P1C&Page=RETR&LANG=Eng&T=301&S=3&O=D Statistics Canada. (2014a). Focus on geography series, 2011 census: census subdivision of Calgary, CY: Alberta. Retrieved from https://www12.statcan.gc.ca/census- recensement/2011/as-sa/fogs-spg/Facts-csd- eng.cfm?Lang=eng&GK=CSD&GC=4806016 Statistics Canada. (2014b). Population projections: Canada, the provinces and territories, 2013 to 2063. Retrieved from http://www.statcan.gc.ca/daily- quotidien/140917/dq140917a-eng.htm Statistics Canada. (2015). Canada's population estimates: Subprovincial areas, July 1, 2014. Retrieved from http://www.statcan.gc.ca/daily- quotidien/150211/dq150211a-eng.htm Stenson, F. (2012). Glenbow Ranch Provincial Park: Grass, hills, and history. Canada: Kingsley Publishing. Sullivan, K. (1988). Ontogeny of time budgets in Yellow-eyed Juncos: Adaptation to ecological constraints. Ecology, 69(1), 118-124. Swihart, R. K., Slade, N. A., & Bergstrom, B. J. (1988). Relating body size to the rate of home range use in mammals. Ecology, 69(2), 393-399. Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3 ed.). New York, New York: Harper Collins College Publishers. Tesky, J. L. (1995). Wildlife species: Canis latrans. Fire effects information system. Retrieved from http://www.fs.fed.us/database/feis/animals/mammal/cala/all.html - BIOLOGICAL%20DATA%20AND%20HABITAT The Association for the Protection of Fur-Bearing Animals. (2013). What is the problem? Retrieved from http://furbearerdefenders.com/what-we-do/living-with- wildlife/coyotes/what-is-the-problem

The City of Calgary. (2014). 2014 civic census results. Retrieved from http://www.calgary.ca/CA/city-clerks/Pages/Election-and-information- services/Civic-Census/2014-Results.aspx The Old Farmer's Almanac. (2015). Moon phases and lunar calendar for Calgary, AB. Retrieved from http://www.almanac.com/astronomy/moon/calendar/ab/calgary/2010-11 Timm, R. M., Baker, R. O., Bennett, J. R., & Coolahan, C. C. (2004). Coyote attacks: An increasing suburban problem. Paper presented at the Proceedings of the Twenty- First Vertebrate Pest Conference, Visalia, California. Town of Cochrane. (2015). Demographics. Retrieved from https://www.cochrane.ca/386/Demographics Tynan, C. T., Ainley, D. G., Barth, J. A., Cowles, T. J., Pierce, S. D., & Spear, L. B. (2005). Cetacean distributions relative to ocean processes in the northern California Current System. Deep Sea Research Part II: Topical Studies in Oceanography, 52(1-2), 145-167. Urban Coyote Research. (2015). Coyote home ranges. Retrieved from http://urbancoyoteresearch.com/coyote-home-ranges Vanpe, C., Gaillard, J.-M., Kjellander, P., Mysterud, A., Magnien, P., Delorme, D., . . . Hewison, A. J. M. (2007). Antler size provides an honest signal of male phenotypic quality in Roe Deer. The American Naturalist, 169(4), 481-493. Venter, O., Brodeur, N. N., Nemiroff, L., Belland, B., Dolinsek, I. J., & Grant, J. W. A. (2006). Threats to endangered species in Canada. BioScience, 56(11). Wang, X., & Tedford, R. H. (2008). Dogs: Their fossil relatives and evolutionary history. New York, New York: Columbia University Press. Weinberg, S. L., & Abramowitz, S. K. (2008). Statistics using SPSS: An integrative approach (2 ed.). Cambridge: Cambridge University Press. White, G. C., Anderson, D. R., Burnham, K. P., & Otis, D. L. (1982). Capture-recapture and removal methods for sampling closed populations (Vol. Report LA-8787- NERP). Los Alamos, New Mexico: Los Alamos National Laboratory. Williams, R. (2012). Comprehensive statistical analysis using SPSS 19.0. Lethbridge, Alberta: Faculty of Health Sciences, University of Lethbridge. Wolff, J. O., & Van Horn, T. (2003). Vigilance and foraging patterns of American elk during the rut in habitats with and without predators. Canadian Journal of Zoology, 81(2), 266-271. Worcester, R. E., & Boelens, R. (2007). The co-existing with coyotes program in Vancouver, B.C. Paper presented at the Wildlife Damage Management Conferences- Proceedings. Wright, G. J., Peterson, R. O., Smith, D. W., & Lemke, T. O. (2006). Selection of northern Yellowstone elk by gray wolves and hunters. Journal of Wildlife Management, 70(4), 1070-1078. Young, S. P., & Jackson, H. H. T. (1951). The clever coyote. Washington, D.C.: Wildlife Management Institute. Zar, J. H. (1999). Biostatistical analysis (4 ed.). Englewood Cliffs, N.J.: Prentice-Hall, Inc.

Appendices

Appendix A Coyote scat is easily identifiable from other animal scat in GRPP. If species was uncertain the scat was rejected. The table below summarizes characteristics of wild animal scat (Halfpenny & Biesiot, 1986, p. 129-148) and domestic animal scat (Lukasik & Alexander, 2012) that could have potentially been found in GRPP. Animal Scat Scat Location Scat Contents Image Description

Coyote and red “Thick cords” “Often several - Insect parts Coyote fox scats will be - Seeds Only the end deposited at the - Grasses that leaves the same point - Bone anus is pointed over the course fragments

even if scat is of several - Hair From (Halfpenny & Biesiot, segmented days” - Feathers 1986). - Teeth “White, gray, May mark their Red fox brown, and territory by black” defecating on “forks in the Red fox: trail, old diameter is carcasses, or usually < 18 prominent From (Halfpenny & Biesiot, millimetres knobs” 1986).

Coyote: diameter is usually between 18-25 millimetres

Domestic dog Diameter can “Tend to Dog food may vary depending defecate in give scat a on the breed grassy areas off “grainy” look trails and Color often has paths” a yellowish hue

“Smooth & tubular”

Cat “Relatively “Deposited on - Insect parts Cougar short tails” elevated points - Seeds (tree stumps - Grasses “Segmented and rocks) near - Bone (broken cords)” the edges of the fragments home range - Hair From (Halfpenny & Biesiot, “Brown, white, and on hunting - Feathers 1986). gray, and black” trails” - Teeth “Near dens, Lynx both scat and urine are deposited in shallow holes and then covered” From (Halfpenny & Biesiot, 1986).

Bobcat

From (Halfpenny & Biesiot, 1986).

Bear “Thick cords May be - Nuts Black bear with blunt ends” deposited on - Berries top of other - “Animal Usually species’ feces protein” deposited in big - Insect parts

piles “When feeding - Grasses for several days - Dandelions “Brown, black, on one carcass, - Thistles and blue-colored will leave - Horsetails scat” many scats around their “A fall diet of day beds” From (Halfpenny & Biesiot, berries will 1986). produce soft to semi-liquid scat”

Weasel “Musky odor” “Badgers - Bone (e.g. Badger) when fresh usually deposit fragments their scat below - Hair “Folded cords ground” - Feathers with long tails often at both ends” From (Halfpenny & Biesiot, “Black and 1986). brown with occasional gray” 79

Raccoon “Short, thick “Tend to - Omnivore even diameter deposit in large diet cords, usually piles of several - Crayfish with flat ends” scats” and on - Fish “branches or “Black to limbs” brown, reddish, or even bleached “Latrines are From (Halfpenny & Biesiot, white” often used 1986). repeatedly” “Characteristic granular appearance”

Skunk “Blunter ends” Striped skunk Same diet as compared to is the only weasels with weasel scat species of the addition of skunk in insects Alberta. It will “make latrines” From (Halfpenny & Biesiot, 1986).

Appendix B Biological reasoning behind the 14 binary multiple logistic regression models. Model Number and Name Why I Included the Variable(s)

1. Global Coyotes tend to avoid humans (George & Crooks, 2006; Porter, 2013; Riley et al., 2003). I extended this observation to all variables related to people. This included vehicles, cyclists and domestic dogs. I included prey because prey is important to coyote survival. Moon phase is related to coyote vocalizations (Bender et al., 1996). Thus, it could potentially affect other aspects of the coyote’s life, such as movement and trail occurrence. Coyote trail occurrence within the past day was included because coyotes are social and can be territorial (Bowen, 1981; Camenzind, 1978; Fox & Papouchis, 2005). Thus, coyotes may want to “investigate” the presence of other coyotes that have used trails. Life cycle stages may also affect movement should coyotes be foraging for offspring, finding a mate or dispersing. Time of day is of interest because coyotes may alter activity around people by becoming more nocturnal (Gehrt et al., 2009; Kitchen et al., 2000; Riley et al., 2003). Season was included because high human use trails may provide more efficient travel in the winter when there is snow cover on trails (Paquet et al., 2010).

2. Natural Factors Coyotes live alongside people in urban areas (Gehrt et al., 2009). Thus, human-related disturbances may not have a significant influence on their occurrence. Consequently, natural factors may be more related to coyote occurrence on trails.

3. Human-Related Disturbances Some articles have demonstrated that coyotes avoid people (e.g. George & Crooks, 2006; Porter, 2013; Riley et al., 2003). Thus, I wanted to explore whether people, as well as other human- related factors (number of domestic dogs, cyclists and vehicles) are strongly related to coyote trail occurrence.

4. Time-Related Studies have analyzed coyote occurrence relative to season (e.g. Porter, 2013) as well as time of day (e.g. George & Crooks, 2006; Riley et al., 2003). I wanted to explore both of these timeframes as well as life cycle stage temporal periods to see if time-related variables are important predictors of coyote occurrence.

5. Animals I was interested in seeing if coyote trail occurrence was potentially related to locating prey as well as avoiding competitors (e.g. conspecifics).

6. Coyote Social Aspects Coyotes are social animals (Fox & Papouchis, 2005). Do social factors affect their movement? Does coyote occurrence on trails the day before affect future coyote occurrence? What about moon phase? Coyotes in groups howl less when there is more moonlight (Bender et al., 1996). Since howling is a form of communication (Mitchell et al., 2006), moon phase is related to sociality. Does it also affect occurrence?

7. People and Dogs Comparing coyote occurrence to people as well as dogs were key goals of the study.

8. People Comparing coyote occurrence to people was the primary goal of the study.

9. Moon Phase Moon phase is related to coyote vocalizations (Bender et al., 1996). Perhaps communication or a lack of communication affects coyote occurrence on trails?

10. Natural Factors + Cyclists I wanted to explore how different human-related 11. Natural Factors + Vehicles disturbances plus natural factors related to coyote occurrence. 12. Natural Factors + Cyclists + Vehicles 13. Natural Factors + People 14. Natural Factors + Dogs