Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Spatiotemporal patterns and reliability of bobcat and Canada lynx occurrence records in British Columbia Gooliaff, TJ 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2018_february_gooliaff_tj.pdf [ 6.1MB ]
Metadata
JSON: 24-1.0362863.json
JSON-LD: 24-1.0362863-ld.json
RDF/XML (Pretty): 24-1.0362863-rdf.xml
RDF/JSON: 24-1.0362863-rdf.json
Turtle: 24-1.0362863-turtle.txt
N-Triples: 24-1.0362863-rdf-ntriples.txt
Original Record: 24-1.0362863-source.json
Full Text
24-1.0362863-fulltext.txt
Citation
24-1.0362863.ris

Full Text

 SPATIOTEMPORAL PATTERNS AND RELIABILITY OF BOBCAT AND CANADA LYNX OCCURRENCE RECORDS IN BRITISH COLUMBIA  by  TJ Gooliaff  BSc, University of the Fraser Valley, 2015  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE  in  THE COLLEGE OF GRADUATE STUDIES (Biology)  THE UNIVERSITY OF BRITISH COLUMBIA OKANAGAN (Okanagan) January 2018 © TJ Gooliaff, 2018 ii The following individuals clarify that they have read, and recommend to the College of Graduate Studies for acceptance, a thesis entitled:  SPATIOTEMPORAL PATTERNS OF BOBCAT AND CANADA LYNX OCCURRENCE RECORDS IN BRITISH COLUMBIA submitted by TJ Gooliaff in partial fulfillment of the requirements of the degree of Master of Science.  Dr. Karen E. Hodges, Irving K. Barber School of Arts and Sciences Supervisor Dr. Jason Pither, Irving K. Barber School of Arts and Sciences Supervisory Committee Member Richard D. Weir, British Columbia Ministry of Environment Supervisory Committee Member Dr. Ross Hickey, Irving K. Barber School of Arts and Sciences University Examiner        iii Abstract   Understanding the distribution of a species is imperative for proper management and conservation to occur, especially in the wake of climate change and other biotic and abiotic changes that are currently occurring across the globe. Numerous species are shifting their ranges in response to these changes; tracking these shifts is the first step in conserving current and future habitats. Provincial or state-wide records, as well as citizen science, can be used to estimate broad-scale distributions, whereas camera trapping has become a common tool for detecting species at smaller scales. However, cryptic and elusive species such as mesocarnivores are difficult to monitor. Here, I use bobcats (Lynx rufus) and Canada lynx (L. canadensis) in British Columbia, Canada, as a case study to examine the data sources available for mapping mesocarnivore distributions, use those data to assess whether bobcat and lynx distributions have undergone any shifts over the past century, and determine whether these very similar-looking species can be reliably identified from camera images. I estimated the distributions of bobcats and lynx in BC using five independent data sources. Trapping records, hunting records, vehicle-kill records, trapper surveys that I distributed, and images that I solicited from the public all indicated that bobcats were restricted to southern BC, whereas lynx occurred throughout most of the interior of the province. I used trapping records and trapper surveys to determine whether bobcats or lynx have shifted their range in BC; these data suggest that distributions have remained stable over the past century. Lastly, I measured agreement among experts in classifying images of bobcats and lynx, and revealed that even experts find it difficult to distinguish between bobcats and lynx from camera images; agreement among experts in their classifications of the images was poor, and experts were inconsistent when asked to reclassify images weeks later. iv Preface  Versions of each of my three data chapters (Chapter 2, Chapter 3, and Chapter 4) have been submitted to journals as separate publications. All three manuscripts are currently in review. I completed all data collection and analyses and wrote each manuscript. The publication arising from Chapter 2 has two additional co-authors: Richard D. Weir and Karen E. Hodges. The publications arising from Chapters 3 and 4 each have one additional co-author: Karen E. Hodges.  Richard Weir provided the trapping records in Chapters 2 and 3, made it possible for me to travel to the regional Ministry of Forests, Lands, and Natural Resource Operations offices to sort through the trapline files and obtain contact information for trappers in Chapters 2 and 3, helped solicit images in Chapter 2, and provided meaningful comments on all three manuscripts. Karen E. Hodges helped think of the research ideas and helped design the studies in Chapters 2, 3, and 4, helped write the trapper survey in Chapters 2 and 3, helped construct the image classification experiment in Chapter 4, and provided meaningful comments on all three manuscripts. Jason Pither also offered useful comments on all three manuscripts.  The University of British Columbia Okanagan  Human Ethics Board approved the survey that I distributed to trappers in Chapters 2 and 3 (project title: trapper survey; certificate # H16-02432), and the online image classification experiment that I distributed to a group of biologists in Chapter 4 (project title: lynx and bobcat photo identification; certificate # H16-03169).   v Table of Contents  Abstract .......................................................................................................................................... iii Preface............................................................................................................................................ iv Table of Contents .............................................................................................................................v List of Tables ............................................................................................................................... viii List of Figures ................................................................................................................................ ix Acknowledgements ....................................................................................................................... xii Dedication .................................................................................................................................... xiv Chapter 1: Introduction ....................................................................................................................1 Chapter 2: Estimating bobcat and Canada lynx distributions in British Columbia from provincial   records and citizen science .............................................................................................6          Literature review and objectives .......................................................................6         Study area ..........................................................................................................8         Methods .............................................................................................................9         Results .............................................................................................................14         Discussion .......................................................................................................18 Chapter 3: Range shifts are not always consistent along range edges: bobcats and Canada lynx    remain stable in British Columbia ................................................................................31         Literature review and objectives .....................................................................31         Methods ...........................................................................................................34         Statistical analysis ...........................................................................................36         Results .............................................................................................................37 vi         Discussion .......................................................................................................39 Chapter 4: Poor agreement among experts in classifying camera images of similar species ........49         Literature review and objectives .....................................................................49         Methods ...........................................................................................................51         Statistical analysis ...........................................................................................54         Results .............................................................................................................56         Discussion .......................................................................................................60 Chapter 5: Conclusions ..................................................................................................................73 References ......................................................................................................................................77 Appendices .....................................................................................................................................91                                       Appendix A: Bobcat and lynx harvest regimes during 1983-2015 in British      Columbia ...................................................................................91                                Appendix B: Wildlife Management Units and registered trapline system in   British Columbia .......................................................................92                                Appendix C: Trapper survey ...........................................................................94                                Appendix D: Wanted poster for bobcat and lynx images ...............................98                                Appendix E: Active traplines during 2000-2013 in British Columbia ...........99                                Appendix F: Elevations of bobcat and lynx images taken during 2008-2017 in      British Columbia separated by season. ...................................100                                Appendix G: Distributions of bobcats and lynx from trapping records     summarized by Management Units in British Columbia each      year during 1983-2013 ............................................................101                                Appendix H: Northern-most confirmed bobcat in British Columbia ...........105 vii                                Appendix I: Online image classification experiment....................................106                       viii List of Tables  Table 2.1.  Details of the data sources used to assess bobcat (Lynx rufus) and lynx (L. canadensis) distributions in British Columbia .......................................................24 Table 2.2.  Number of Management Units containing records of bobcats (Lynx rufus) and lynx (L. canadensis) from each data source during 2008-2017 in British  Columbia ................................................................................................................25 Table 3.1.  Coincidence summaries for the distributions of bobcats (Lynx rufus) and lynx (L. canadensis) summarized by Management Units in British Columbia at two scales: provincial (1983-2013) and central BC (1935-2013) .............................................43 Table 3.2.  Changes noticed by trappers (n = 138) in the presence, elevation, and abundance of bobcats (Lynx rufus) and lynx (L. canadensis) on their traplines in British Columbia ................................................................................................................44 Table 3.3.  Correlates of bobcat (Lynx rufus) and lynx (L. canadensis) trapper harvest during 1983-2009 in British Columbia ..............................................................................45 Table 4.1.  Characteristics of the 15 image categories within the six trials. ............................65 Table 4.2.  Examples of images with poor agreement among experts in their classifications (n = 27 experts) ...........................................................................................................66 Table 4.3.  Agreement among all experts (n = 27) in their classifications of images within each category of images .........................................................................................67 Table A.1.  Bobcat (Lynx rufus) and lynx (L. canadensis) hunting and trapping seasons, bag limits, and compulsory inspection and reporting requirements during 1983-2015 in British Columbia ................................................................................................91 ix List of Figures  Figure 2.1.  Distribution of bobcat (Lynx rufus) a) trapping records (2008-2013), b) compulsory records from hunting (2008-2015), c) presence and absence reported from trapper surveys (2008-2016), d) vehicle-kill records (2008-2016), and e) images submitted by the public (2008-2017) in British Columbia ........................26 Figure 2.2.  Distribution of lynx (Lynx canadensis) a) trapping records (2008-2013), b) compulsory records from hunting (2008-2015), c) presence and absence reported from trapper surveys (2008-2016), d) vehicle-kill records (2008-2016), and e) images submitted by the public (2008-2017) in British Columbia ........................27 Figure 2.3.  Elevations of bobcat (Lynx rufus; black bars) and lynx (L. canadensis; gray bars) images taken during 2008-2017 in British Columbia estimated from a) all images, and b) only images from Management Units that contained records of both species from ≥1 data sources (i.e., areas of range overlap between bobcats and  lynx) .......................................................................................................................28 Figure 2.4.  Bobcat (Lynx rufus) and lynx (L. canadensis) distributions in British Columbia during 2008-2017 estimated from trapping records, compulsory records from hunting, trapper surveys, vehicle-kill records, and images submitted by the  public ......................................................................................................................29 Figure 2.5.  Range overlap between bobcats (Lynx rufus) and lynx in (L. canadensis) British Columbia during 2008-2017 ..................................................................................30 Figure 3.1.  Distributions of bobcats (Lynx rufus) and lynx (L. canadensis) from trapping records summarized by Management Units in British Columbia in a) 1983-1993, x b) 1994-2003, and c) 2004-2013 ............................................................................46 Figure 3.2.  Distributions of bobcats (Lynx rufus) and lynx (L. canadensis) from trapping records summarized by Management Units in central British Columbia in a) 1935-1961, b) 1962-1987, and c) 1988-2013 ..................................................................47 Figure 3.3.  Total harvests (bars and left y-axis) and average pelt prices (line and right y-axis) of bobcats (Lynx rufus; top) and lynx (L. canadensis; bottom) from 1983 to 2013 in British Columbia ................................................................................................48 Figure 4.1.  Images of bobcats (Lynx rufus; white circles; n = 805) and lynx (L. canadensis; black circles; n = 807) taken during 2008-2017 .....................................................68 Figure 4.2.  Distribution of a) the number of experts that classified individual images as the majority classification and b) the proportion of agreement scores among all 27 experts for individual images in all categories excluding the 40 images for which I provided locations (n = 259 images) ......................................................................69 Figure 4.3.  Agreement among all experts (n = 27) in their classifications of images within each category of images .........................................................................................70 Figure 4.4.  Examples of how the location of an image and the visible features of an animal can affect expert classification ...............................................................................71 Figure 4.5.  Average probability that the majority classification of a randomly selected subset of experts matched the majority classification of all 27 experts, calculated from all images excluding the 40 images for which I provided locations   (n = 259 images) ....................................................................................................72 Figure B.1.  Wildlife Management Units in British Columbia ..................................................92 Figure B.2.  Registered trapline system in British Columbia .....................................................93 xi Figure E.1.  Active traplines during 2000-2013 in British Columbia ........................................99 Figure F.1.  Elevations of bobcat (Lynx rufus; black bars) and lynx (L. canadensis; gray bars) images taken during 2008-2017 in British Columbia estimated from all images taken during a) April 1 - September 30, and b) October 1 - March 31 ................100 Figure G.1.  Distribution of bobcats (Lynx rufus) from trapping records summarized by Management Units in British Columbia each year during 1983-2013 .................101 Figure G.2.  Distribution of lynx (Lynx canadensis) from trapping records summarized by Management Units in British Columbia each year during 1983-2013 .................103 Figure H.1.  Northern-most confirmed bobcat (Lynx rufus) in British Columbia ....................105 Figure I.2.  Typical screen shot of one of the image pages in one of the six online image classification trials ................................................................................................106             xii Acknowledgements   First and foremost I would like to thank my family and my fiancé Hailey Deptuck for their overwhelming support during this endeavor - I could not have reached this achievement without you.  My gratitude extends to the faculty and students of the University of British Columbia Okanagan who helped make my time in grad school the best couple years of my life. I would like to thank my supervisor, Karen Hodges, for giving me this opportunity and putting me in a position to succeed in this degree and beyond. She provided excellent guidance and supported me from start to finish. My committee members, Jason Pither and Richard Weir, were also invaluable resources and greatly improved this work. My lab mates and fellow students made coming to campus something that I looked forward to and something that I will truly miss. The UBCO Student Chapter of The Wildlife Society, the Vertebrate Conservation Discussion Group, and being a teaching assistant for Tristyn Hay and Robert Lalonde were particular highlights of my time at UBCO.  This work would not have been possible without generous contributions from many people. British Columbia’s Ministry of Forests, Lands, and Natural Resource Operations provided harvest records and spatial data for Wildlife Management Units and traplines. BC’s Ministry of Transportation and Infrastructure provided vehicle-kill records. Regional wildlife biologists helped me sort through physical trapline files and were always available to answer my countless questions.  I would like to thank the BC Trapper’s Association for welcoming me into the trapping community and supporting my research by publishing articles in the BC Trapper Magazine and xiii giving me the opportunity to speak at their Convention and local meetings. I would like to thank everyone that took the time to complete my trapper survey; the optional comments box was always overflowing! I met a great number of trappers who took the time to share their knowledge with me, many of which I now call my friends. I would like to thank Pete Wise and Ron Lancour for being especially helpful and for teaching me about the world of trapping.  The response to my call for bobcat (Lynx rufus) and Canada lynx (L. canadensis) images from the public was overwhelming. The enormous number of images that I received was far greater than I could have imagined; I would like to thank everyone who submitted images and sighting information, especially Parks Canada Agency and BC Parks. I would also like to thank all of the media reporters and journalists that published stories about my research and my search for images, all hunting and outdoor stores across BC that displayed my ‘wanted’ poster, Richard Weir for publishing my notice in the 2016-2018 BC Hunting and Trapping Synopsis, and everyone else who helped spread the word about my project. The interest from the public was no doubt because of the great coverage that my research received.   Finally, I would like to thank the 27 biologists who took the time to participate in my image classification experiment. I greatly appreciate all of the hours that everyone spent looking at hundreds of bobcat and lynx images. The large number of willing participants turned this fun experiment into a genuine contribution to the camera trapping literature.  Funding for this work was provided by a Natural Sciences and Engineering Research Council of Canada research grant awarded to Karen Hodges, as well as UBCO fellowships awarded to me. I was also lucky enough to obtain the 2015 BC Conservation Foundation John B. Holdstock Scholarship and the 2016 BCTA Bob Gibbard Bursary.  xiv    In memory of my grandparents  George York (1928 -2011) Louise York (1950 - 2016)                    1 Chapter 1  Introduction   Studying and monitoring mesocarnivores is a difficult task; mesocarnivores are often rare, elusive, and avoid humans (Gese 2001, Long et al. 2008, Day et al. 2016). Dauntingly, researchers and managers are often tasked with mapping broad-scale distributions of mesocarnivores to obtain baseline information about the species (US Fish and Wildlife Service 2005, Interagency Lynx Biology Team 2013). Understanding the distribution of a species can help define research questions and prioritize conservation and management. Such broad-scale distribution studies are limited to using province or state-wide data sources such as harvest records (Litvaitis et al. 2006, Golden et al. 2007, Lavoie et al. 2009, Robichaud and Boyce 2010) and vehicle-kill records (Woolf et al. 2002, Brockie et al. 2009), or using citizen science (Palma et al. 1999, Gese 2001, Reed et al. 2017).  Tracking the distributions of mesocarnivores is especially important under climate change and other biotic and abiotic changes, which are causing numerous species across the planet to shift their ranges (Parmesan and Yohe 2003, Chen et al. 2011). Range shifts are defined as a change in the distribution of a species over time (Lenoir and Svenning 2013), and are most noticeable at range peripheries (Gaston 2009). Most documented range shifts have occurred at northern latitudes between 30°N and 60°N, and usually result in leading-edge expansion or trailing-edge contraction (Lenoir and Svenning 2015). Chen et al. (2011) reported recent range shifts for terrestrial species in the northern hemisphere to be northward at a median rate of 16.9 km per decade, and to higher elevations at a median rate of 11.0 m per decade.  2  In addition to assessing broad-scale distributions, more detailed detections at finer spatial and temporal scales are often of interest for evaluating habitat selection, estimating occupancy, and obtaining other ecological information for mesocarnivores, for which camera trapping (i.e., the use of remote cameras) has become a common tool (Rowcliffe and Carbone 2008, O’Connell et al. 2011). Such studies collect enormous numbers of wildlife images that must be classified based on the species that they contain (He et al. 2016). Because correct species classification is crucial (Royle and Link 2006, Miller et al. 2011, Molinari-Jobin et al. 2012, Costa et al. 2015), experts usually classify the images (McShea et al. 2016), although even experts can misclassify images (Gibbon et al. 2015, Austen et al. 2016, Swanson et al. 2016). Numerous species, including many mesocarnivores such as bobcats (Lynx rufus) and Canada lynx (L. canadensis; hereafter lynx), are difficult to tell apart, making image classification challenging (McShea et al. 2016, Swanson et al. 2016).  Bobcats and lynx are closely related and similar-looking mesocarnivores, but their ecology is very different. Bobcats are habitat and dietary generalists; they occur in a wide range of habitats including deserts, grasslands, wetlands, and coniferous forests (Hansen 2007). Their diet includes many species including deer (Odocoileus spp.), snowshoe hare (Lepus americanus), eastern cottontail (Sylvilagus floridanus), North American red squirrel (Tamiasciurus hudsonicus), as well as rodents and birds (Hansen 2007, Newbury 2013). Bobcats are extremely adaptable and are the most wide-spread wild felid in North America (Anderson and Lovallo 2003); they are common across most of the contiguous United States, Mexico, and southern Canada (Hansen 2007, Roberts and Crimmins 2010).  In contrast, lynx are confined to the boreal forest across Canada and Alaska, as well as the mountain ranges that extend into the northern contiguous US (McKelvey et al. 2000, Mowat  3 et al. 2000, Lewis 2016). Lynx are specialists, and their geographic distribution is concordant with that of their primary prey, snowshoe hares (Mowat et al. 2000). In local habitats, lynx prefer areas with a high abundance of snowshoe hares (Aubry et al. 2000). Lynx and snowshoe hares exhibit a well-documented 10-year population cycle, in which snowshoe hare populations continuously rise and fall, followed by lynx with a slight lag (Elton and Nicholson 1942, Krebs et al. 2001). The cycle is most pronounced in northern Canada and Alaska, but still occurs in southern populations at lower amplitudes (Hodges 2000a, b).  Bobcats and lynx are similar in size and stature, but have slight anatomical differences. The average mass of bobcats is lower than lynx (9-14 kg compared to 9-15 kg for males, and 6-10 kg compared to 7-12 kg for females; Hatler and Beal 2003, Hatler et al. 2003). However, bobcats can reach larger sizes than lynx (Buskirk et al. 2000), and are commonly larger than lynx in areas where their ranges overlap (Hansen 2007). Lynx have long legs and large snowshoe-like paws, making them well adapted for traveling across deep snow (Murray and Boutin 1991, Hoving et al. 203, Pozzangera et al. 2016). In contrast, bobcats have short legs and small paws that have double the foot-loading of lynx (Parker et al. 1983, Buskirk et al. 2000). Lynx have more pronounced facial ruffs and longer ear-tufts, as well as shorter, solid black-tipped tails, as opposed to the longer, black and white-tipped tails of bobcats. Bobcats also have black heel marks that are absent on lynx, and usually have more brownish and spotted fur coats compared to the grey-silver coats of lynx.  In British Columbia, Canada, bobcats and lynx are managed as furbearers and are legally harvested through hunting and trapping seasons (BC Ministry of Forests, Lands, and Natural Resource Operations 2017). Bobcats are common and are not at risk of extinction in Canada or the US, whereas lynx are listed as Threatened in the contiguous US but not in Canada (US Fish  4 and Wildlife Service 2000). Bobcats and lynx are highly valued by trappers and naturalists alike (Hatler et al. 2003, Hatler and Beal 2003). Ecologically, both species play important roles in the ecosystems that they inhabit. However, their provincial distributions are poorly known in BC. Further, it is unclear if their distributions may be changing in the face of climate change and other biotic or abiotic changes in this province.  Here, I use bobcats and lynx in BC as a case study to examine the data sources available for mapping mesocarnivore distributions, and use those data to examine potential range shifts. Because bobcats and lynx are similar-looking species, I also measure agreement among experts in classifying these species from camera images. My objectives were to 1) estimate the current provincial distributions of bobcats and lynx in BC, 2) determine whether the distribution of each species has shifted over the past century, and 3) examine whether bobcats and lynx can be reliably classified from camera images.  In Chapter 2, I examined the data sources that exist in BC for assessing mesocarnivore distributions. I compiled five independent sources of bobcat and lynx records in BC to gain a better understanding of their provincial distributions: trapping records, hunting records, trapper surveys, vehicle-kill records, and images that I solicited from the public. I analyzed each of these data sources separately to compare resulting distributions from the different records, and then I combined these data sources together to provide the best composite estimate of current bobcat and lynx distributions in BC.  After determining the current distribution of each species, I then examined whether those distributions have shifted over the past century in Chapter 3. I used trapping records to create a series of distribution maps for each species for different time periods, and then I compared those distributions over time to see if they have changed. I also surveyed trappers across BC to  5 determine whether trappers have noticed any evidence of range shifts, either in latitude or elevation.  In Chapter 4, I took a closer look at one of the other data sources, images that I solicited from the public, to determine whether image classifications in camera trapping studies are reliable for such similar-looking species. I asked experts to classify a subset of images to test whether the season, background habitat, time of day, and the visible features of each animal (e.g., face, legs, tail) affected agreement among them about the species, whether knowing the location of an image affected agreement in their classifications of images, and whether experts were consistent in their classifications when asked to reclassify the same images weeks later.  Lastly, in Chapter 5, I synthesize my findings. I detail the answers to each of my objectives, and discuss the implications for mesocarnivore research and conservation.              6 Chapter 2  Estimating bobcat and Canada lynx distributions in British Columbia from provincial records and citizen science  Literature review and objectives  Understanding the distribution of a species is crucial for wildlife researchers and managers (US Fish and Wildlife Service 2005). For example, classifying core and peripheral populations can help prioritize the scope and scale of management actions and identify knowledge gaps that hamper effective management (Interagency Lynx Biology Team 2013). Distribution estimates also provide insights into the ecology of a species, such as broad-scale habitat associations (Aubry et al. 2007). For these reasons, biologists are often tasked with estimating the distribution of a species, which requires many reliable occurrence records across large spatial scales within a narrow time frame.  Broad distribution studies, such as those for entire states or provinces, are often limited to using occurrence data such as harvest records and vehicle-kill records (Gese 2001). While harvest records should be used with caution when assessing abundance or population trends (Poole and Mowat 2001, DeVink et al. 2011, Dorendorf et al. 2016), they are effective at evaluating distribution (Litvaitis et al. 2006, Golden et al. 2007, Lavoie et al. 2009, Robichaud and Boyce 2010). In jurisdictions that have standardized vehicle-kill reporting systems, these data share many of the same advantages as harvest records and can aid in assessing distribution (Woolf et al. 2002, Sielecki 2004, 2005, Brockie et al. 2009).  Another approach is citizen science, such as sending surveys to people with local species  7 knowledge like trappers (Bridger et al. 2016), or soliciting public sighting reports (Palma et al. 1999, Reed et al. 2017). However, images or physical evidence must accompany public sighting reports to be verifiable (Aubry and Jagger 2006, McKelvey et al. 2008, Garrote and Ayala 2015, Roy et al. 2016). While there are concerns over the standardization and reliability of data collected using citizen science (Kosmala et al. 2016), the volume of data, and their spatial and temporal coverage, cannot be obtained by any other method. However, obtaining enough records to estimate distribution is a challenge for cryptic and elusive mesocarnivores (Gese 2001, Long et al. 2008, Day et al. 2016) such as bobcats (Lynx rufus) and Canada lynx (L. canadensis; hereafter lynx).  Bobcats and lynx are congeneric felids with wide distributions across North America, but they are largely allopatric at the continental scale (Koen et al. 2014). Bobcats occur throughout the contiguous United States with their range extending from Mexico to southern Canada (Anderson and Lovallo 2003, Hansen 2007). In contrast, lynx occur throughout most of Canada and Alaska, as well as the mountain chains that extend into the contiguous US (McKelvey et al. 2000, Lewis 2016). British Columbia, Canada, contains the northwestern range margin of bobcats and is close to the southwestern range margin of lynx (Hatler and Beal 2003, Hatler et al. 2003). However, current provincial distributions of both species are poorly known; published distribution maps from various time periods are inconsistent with each other (Woolf and Hubert 1998, Anderson and Lovallo 2003, Hansen 2007, Peers et al. 2013, Koen et al. 2014).  In particular, it is unknown where in BC the two species overlap, making it difficult to frame questions about behaviour and ecology of these species in sympatry. Identifying these areas is of interest because of the current debate surrounding interspecific competition between these species. Bobcats may exhibit interference and exploitation competition over lynx (Buskirk  8 et al. 2000), and bobcats may outcompete lynx at broad scales (Parker et al. 1983, Hoving et al. 2003). Peers et al. (2013) suggested that in areas of broad-scale sympatry lynx reduce their habitat niche breadth, while bobcats increase theirs. An additional conservation concern is hybridization between these species (Murray et al. 2008), which has been sporadically detected in eastern North America and near the Great Lakes, but not in western North America (Schwartz et al 2004, Homyack et al. 2008, Koen et al. 2014).  Here, I combine five independent sources of bobcat and lynx records in BC. My objectives were to 1) compare bobcat and lynx distributions derived from each of the five methods, and 2) combine all data together to provide reliable estimates of the distribution of each species in BC during the past 10 years (2008-2017). I show the northern range limit of bobcats, and the degree of range overlap between bobcats and lynx. I also highlight the advantages and limitations of each data source for estimating mesocarnivore distributions.  Study area  BC contains a rich diversity of habitats including desert, temperate rainforest, boreal forest, and alpine (BC Ministry of Forests and Range 1998). In BC, bobcats and lynx are managed as big game species within Wildlife Management Units and as furbearers on a registered trapline system (BC Ministry of Forests, Lands, and Natural Resource Operations 2017; see Appendix A for the harvest regimes of bobcats and lynx). The province is divided into 225 Management Units ranging from 465 km² to 18,982 km² (   = 4,216 km²), and currently contains 2,451 traplines ranging from 0.6 km² to 23,310 km² (   = 366 km²; see Appendix B for maps of BC’s Management Units and registered traplines). Traplines cover most of BC except for urban centers and some provincial and national parks. Management Units and traplines  9 overlap and do not always share common boundaries.  Methods  I assembled data from five sources of bobcat and lynx records in BC, which varied in the number of records that they contained as well as their spatial and temporal resolutions (Table 2.1). I combined data from the past 10 years (2008-2017) to map current distributions of each species; this duration accommodates any cyclicity, as well as ensuring that I had reasonable numbers of data points from the different records. Trapping records indicated that bobcat and lynx distributions have remained stable over the past century in BC (Chapter 3), thus broad-scale distributions have not changed during my 10-year period.  Records of all mesocarnivores that are trapped or hunted in BC are maintained by the Ministry of Forests, Lands, and Natural Resource Operations. Whenever a licensed trapper sells a pelt to a fur buyer (including taxidermists), the fur buyer is required to collect and remit a royalty to the Province along with a report detailing the Crown trapline (area-based Crown-land tenure for which the trapper has exclusive right to harvest furbearing animals), or, if the harvest was on private land, the Management Unit from which the harvest occurred (BC Ministry of Forests, Lands, and Natural Resource Operations 2017). All hunters are required to have their kills inspected by certified personnel (i.e., compulsory inspection) or must self-report their harvest (i.e., compulsory reporting), whereas trappers are required to compulsory report their kills only in certain regions (BC Ministry of Forests, Lands, and Natural Resource Operations 2017).  I obtained the most recent fur royalty reports (2008-2013), compulsory inspection and compulsory reporting records (2008-2015), and spatial data for registered traplines and  10 Management Units from BC’s Ministry of Forests, Lands, and Natural Resource Operations. I removed six anomalous trapping records (five bobcats in northern BC and one lynx on the southern coast) that I considered as highly unlikely based on distributions from the other data sources. Those records are almost certainly errors because there is no trapping season for bobcats in northern BC or for lynx on the southern coast, thus they cannot be sold to fur buyers and instead must be turned over to the Province (BC Ministry of Forests, Lands, and Natural Resource Operations 2017). I called the trapper for one of those bobcat records and confirmed it was a mistake. Contact information was not available for the trappers of the remaining five records.  I combined compulsory inspection and compulsory reporting records from all hunted animals. I did not include compulsory records from trapping because it was required in only some regions. I cross-referenced trapping records with the trapline files at each regional Ministry of Forests, Lands, and Natural Resource Operations office to obtain a history of trapline amalgamations. Amalgamation occurred when two adjacent traplines owned by the same trapper were joined into one trapline. Approximately 13% of traplines reported in the trapping records had amalgamated since the registered trapline system was established in 1926. I compiled harvest totals from amalgamated traplines to report all harvest on traplines as they existed in 2013. Then I compiled records (from Crown traplines and private land) by Management Unit by identifying the Management Unit that contained most of each trapline.  I summarized trapping records and compulsory records within Management Units in a Geographic Information System, which provided a clear spatial resolution at the provincial scale. I considered only Management Units that were active during the monitoring period. For trapping records, I considered Management Units to be active if they contained at least one ‘active  11 trapline’, which I defined as those where the trapper had harvested ≥1 bobcat, lynx, red fox (Vulpes vulpes), coyote (Canis latrans), wolf (C. lupus), or wolverine (Gulo gulo) during the monitoring period. This screen implies that there were traps that could have captured bobcats or lynx in those locations and times (e.g., traps for marten (Martes americana or M. caurina) and beaver (Castor canadensis) would not catch bobcats or lynx). For compulsory records, I defined ‘active Management Units’ as those having an open hunting season for that species.  In addition to analyzing harvest records, I mailed surveys to owners of traplines that harvested any species in ≥5 years during 2000-2013 and harvested ≥1 bobcat, lynx, red fox, coyote, wolf, or wolverine during that time (see Appendix C for the survey). I obtained contact information for trappers from BC’s Ministry of Forests, Lands, and Natural Resource Operations. My survey obtained ethics approval from the University of British Columbia (certificate # H16-02432). I requested that completed surveys be mailed back, but I also provided the option of completing an online version. To encourage participation, in 2016 I presented at trapper meetings and published two articles in the BC Trapper magazine.  I asked trappers to assess the presence of bobcats and lynx based not only on harvest totals, but also on sightings and sign (e.g., tracks, scat). On the survey, I asked if bobcats and lynx were present on their trapline each year during 2000-2016. If trappers indicated that bobcats or lynx were present any year between 2008 and 2016, I classified that species as present during the monitoring period. I summarized this information within Management Units. If there were multiple responses for the same Management Unit, I assigned the species as present if any response showed presence. In addition, I asked trappers if bobcats were generally at higher, lower, or the same elevations as lynx on their traplines.  12  I also obtained the most recent vehicle-kill records for bobcats and lynx (2008-2016) from BC’s Ministry of Transportation and Infrastructure (2017) using their Wildlife Accident Reporting System data use license agreement. As part of the Wildlife Accident Reporting System, daily patrols are conducted along all provincial highways to collect carcasses and detailed information about the species and location of vehicle-killed animals (Sielecki 2004, 2005). Despite the limitations that only 25-35% of wildlife mortalities are reported, and maintenance staff responsible for reporting vehicle-killed animals are not experts in wildlife identification (Sielecki 2005), these data are collected consistently across the entire province and provide carcass data independent from harvest records. I assigned geographic coordinates based on locations reported in the database. I mapped all records as point locations along provincial highways and then summarized all records within Management Units.  Finally, between November 2015 and March 2017, I solicited images of bobcats and lynx from the public across BC. I advertised on social media, distributed ‘wanted’ posters to outdoor stores (see Appendix D for the wanted poster), and published notices in the 2016-2017 BC Hunting and Trapping Regulations Synopsis, sportsman magazines, and naturalist society newsletters. I contacted provincial government biologists, Parks Canada Agency, and private contractors. I also conducted three television, seven radio, and seven newspaper interviews. In addition, regional and provincial newspapers published ~46 articles about my search for images.  I mapped the images that were taken during 2008-2017. I assigned geographic coordinates based on location information provided by the submitters. However, such location descriptions varied greatly in quality, ranging from the estimated distance along a road to geographic coordinates. I used Google Earth to identify roads and landmarks, measure distances, and estimate image locations. If submitters were vague in their location descriptions, or I could  13 not find their indicated locations, I contacted those people for clarification. I mapped images as point locations, and then summarized all records within Management Units.  I independently classified each image as ‘bobcat’, ‘lynx’, or ‘unknown’, rather than relying on what the submitters stated. In a separate study, I asked 27 experts to classify a subset of images; my classifications matched 286 of the 288 images with a majority classification of ‘bobcat’ or ‘lynx’, supporting my ability to distinguish between these two species from images (Chapter 4). Furthermore, 63% of submissions included several images of the same animal, which improved my classifications because I saw multiple angles of those animals that showed different features (e.g., facial ruff, ear tufts, paws). However, bobcats and lynx are difficult to tell apart (Chapter 4), thus I recognize that my image classifications may have some inaccuracies.   These data sources did not lend themselves to detailed distribution modelling or analyses using mixed models. Harvest records contained no measure of effort, making it impossible to interpret whether bobcats or lynx were not trapped or hunted in certain areas because they were absent, or because they were not being targeted (Poole and Mowat 2001, Dorendorf et al. 2016). Furthermore, harvest totals are correlated with many environmental and socioeconomic factors (Gerht et al. 2002, Hiller et al. 2011, Kapfer and Potts 2012). Thus, with no measure of effort, harvest records are presence-only data and by themselves cannot be used to assess absence or abundance (Erickson 1982, DeVink et al. 2011, Dorendorf et al. 2016). The other data sources were also not amenable to distribution modelling or mixed models; trapper surveys had low spatial resolution, vehicle-kill records were few and were limited to provincial highways, and images submitted by the public were biased towards roads and towns, and contained low-precision location information, thus fine-scale analyses of image locations would not be reliable.   14 Results  During 2008-2015, harvest totals varied greatly between trapping records and compulsory records from hunting, but they displayed similar patterns for each species (Figure 2.1, Figure 2.2). Of all traplines in BC, only 12-18% (   = 15%) were active in any one year, but these were spread across the province. During 2008-2013, 99-309 bobcats (   = 205.5 ± 37.61) and 655-1,704 lynx (   = 1,077.0 ± 159.16) were trapped annually. Only 82 traplines reported harvest of both species, accounting for 13% of provincial traplines that had harvested either bobcats or lynx. During 2008-2015, 5-57 bobcats (   = 33.0 ± 5.89) and 13-57 lynx (   = 27.8 ± 6.08) were hunted and then compulsory inspected or reported annually.  I documented 796 (33%) traplines with harvest of bobcats, lynx, or other species for which similar-sized traps were used (i.e., red fox, coyote, wolf or wolverine). I distributed surveys to 675 traplines for which correct mailing addresses were available, and I received 157 completed surveys (23% response rate). Those trappers were spread across BC, although there were large areas with no respondents (see Appendix E for a map of active traplines and those that responded to my survey). Bobcat and lynx distributions estimated by trapper’s observations matched the harvest distribution of each species (Figure 2.1, Figure 2.2). Both species occurred on 41 (26%) of the 157 traplines with responding trappers, and of these, 76% (n = 31) of trappers reported that bobcats were generally at lower elevations than lynx, whereas 10% (n = 4) reported that both species were generally at the same elevations, and 14% (n = 6) were unsure. No trappers reported that bobcats generally occurred at higher elevations than lynx.  Few vehicle-kill records existed for bobcats and lynx during 2008-2016. Each year, 3-10 bobcats (   = 5.8 ± 0.70) and 0-5 lynx (   = 2.5 ± 0.72) were reported dead along provincial highways. In total, 52 bobcats were reported across southern BC and 15 lynx were reported  15 across the interior of the province, which was consistent with the distributions derived from other records (Figure 2.1, Figure 2.2).  Finally, in response to my public solicitation for images, I received 4,399 images and 141 videos comprising 1,674 separate detections from 695 unique submitters (71% of detections were submitted by the public, 23% by the provincial government, 4% by Parks Canada Agency, and 2% by private contractors). Of these images and videos, 4,444 were taken during 2008-2017, comprising 1,635 separate detections during that time period (805 bobcats, 807 lynx, 7 other species, and 16 unknown). Submissions contained a median of two images per sequence (range = 1-38), and sources ranged from remote camera images to opportunistic recordings. Images were submitted via email throughout my study, although submissions spiked whenever and wherever my project received media coverage. Images were submitted from across the province except for remote areas, although most images were taken close to roads and towns. Many (33%) images were not classified by their submitter, thus I suspect that people were unsure of the correct species classifications. Most (67%) images were classified by their submitter, but in a few cases I think that bobcats and lynx were mistaken for each other, with 3% (n = 27) of those bobcat images misclassified as ‘lynx’, and 3% (n = 28) of those lynx images misclassified as ‘bobcats’. Further, one cougar (Puma concolor) image was misclassified as ‘lynx’, whereas one cougar, one coyote, and three house cat images were misclassified as ‘bobcats’.  Image detections of each species matched their distributions indicated by the other record types (Figure 2.1, Figure 2.2). Bobcats and lynx were detected from images in 67% and 61% of Management Units with carcass data (i.e., trapping records, compulsory records, or vehicle-kill records), respectively. Bobcats and lynx were detected from images in only seven and eight Management Units that had no carcasses, respectively.  16  Image detections of bobcats generally occurred at lower elevations than lynx. However, I stress that the elevations of individual records are not precise due to my having to estimate image locations from often vague descriptions provided by the submitters; 66% of locations provided with images were geographic coordinates or residential addresses, while 34% had less detailed descriptions. Thus, the following elevation values are coarse estimates. Most (90%) bobcat detections were <1,112 m in elevation (median of all bobcat images = 516 m, sd = 396 m, n = 805; Figure 2.3a), whereas 90% of lynx detections were >696 m (median of all lynx images = 1,041 m, sd = 407 m, n = 807; Figure 2.3a). Considering only areas where bobcats and lynx overlapped, the same pattern occurred; bobcat detections generally occurred at lower elevations than lynx in Management Units that contained records of both species from ≥1 data sources (median of these bobcat images = 664 m, sd = 338 m, n = 516; median of these lynx images = 1,339 m, sd = 364 m, n = 456; Figure 2.3b). I documented two bobcats and 16 lynx at elevations >2,000 m, all in southeastern BC. Both bobcat detections were during September, whereas the lynx detections occurred between March and December (see Appendix F for graphs showing the elevations of image detections separated by season). The highest bobcat record was 2,149 m whereas for lynx it was 2,271 m, both in Kootenay National Park. In the southern interior, eight remote cameras detected both species, at elevations between 550 m and 1,780 m. However, while both species were detected during the same season in five of those instances, bobcats and lynx were never detected during the same month.  Combining all data sources together provides my best estimate of current bobcat and lynx distributions (Figure 2.4). In total, I compiled 2,415 bobcat records and 7,631 lynx records from the five data sources. Locations of detection records from all data sources were spatially consistent at the Management Unit scale (Table 2.2). Bobcats and lynx were detected in 112  17 (50%) and 164 (73%) Management Units, respectively. The number and kind of records do not reflect abundance. For example, some (n = 5) remote northern Management Units contained no lynx records; lynx were likely present, but there were few people to detect them (Figure 2.4).  At the scale of Management Units, the total area of bobcat distribution in BC was 300,049 km2 (32% of province), whereas the total area of lynx distribution was 744,664 km2 (79% of province). During 2008-2017, bobcat records occurred across the southern half of BC whereas lynx records occurred throughout most of the interior of the province. Neither species was reported from Vancouver Island (31,285 km2), Haida Gwaii (10,180 km2), or any of the smaller coastal islands. Bobcat records were most common in the southern quarter of the province, but bobcat records were also present in central BC. Lynx records were most common in the central interior and northeastern BC. Lynx records were absent from the southern coast and rare throughout mountainous regions of the province. Bobcat and lynx records overlapped in much of southern BC at broad-scale (Figure 2.5). Bobcat records occurred in 52% of Management Units that had lynx records, whereas lynx records occurred in 76% of Management Units that had bobcat records.  The northern range limit of bobcats is approximated by Highway 16 (~54.0° N), which runs west to east across BC from Prince Rupert to Mount Robson Provincial Park. The most northerly bobcat record that I obtained was near Houston, BC (~54.4° N; see Appendix G for images of this bobcat). This bobcat was trapped in 2015 and turned over to the Province because there was no trapping season for bobcats in this Region. I photographed this bobcat and included it in my image records. This bobcat was female (K. Dixon, BC Ministry of Forests, Lands, and Natural Resource Operations, personal communication), and DNA sequencing confirms she was pure bobcat (R. Weir, BC Ministry of Environment, unpublished data). To my knowledge, this  18 record is the most northerly verified bobcat reported in BC.  Discussion  Until now, bobcat and lynx distributions in BC were poorly described. I combined records from five independent data sources to estimate the current provincial distribution of each species. While each method had unique advantages and limitations, distributions were largely congruent across the five data sources, showing bobcats in the southern half of BC, and lynx across most of the interior of the province.  I show that the northern range limit of bobcats is just south of Highway 16, which runs west to east at latitudes between ~52° N and ~55° N), close to the northern boundary of Management Units for which there is an open season for trapping and hunting bobcats. If only harvest records were considered, it would be impossible to conclude whether this reported range limit was simply an artifact of administrative boundaries. However, the trapper surveys along this potential range limit reported that bobcats were absent, and I received hundreds of lynx images in this area but no bobcat images (Figure 2.1, Figure 2.2). Therefore I infer that there is currently no established bobcat population north of Highway 16. These results strongly suggest that previously published distribution maps overestimate the current and historical northern range limit of bobcats (Woolf and Hubert 1998, Anderson and Lovallo 2003, Hansen 2007, Peers et al. 2013). Based on province-wide trapping records dating back to 1983, I am confident that bobcats have not had sustainable populations as far north as these sources reported (Chapter 3).  I also show that bobcats and lynx have broadly overlapping distributions in southern BC at the scale of Management Units, but at finer scales such as traplines, bobcat and lynx co-occurrence appears to be rare. Few traplines harvested both species (n = 82). On traplines that  19 did contain both species, most (76%) trappers reported that bobcats were generally found at lower elevations than lynx. Furthermore, images submitted by the public varied greatly in elevation between bobcats and lynx. The number of bobcat detections decreased from 0 m to 2,000 m, whereas the number of lynx detections resembled a normal distribution centered around 800-1,200 m. Elevations at which the two species most overlapped were between 400-1,200 m; below 400 m, lynx detections were rare, and higher than 1,200 m, bobcat detections were rare. However, there were a few (n = 8) detections of both species by the same remote camera, but never within the same month. Thus, although none of my data sources were designed to detect fine-scale overlap between these species, these patterns suggest that bobcats and lynx may be spatially or temporally segregated by elevation in individual landscapes.  Spatial separation of bobcats and lynx may be partly explained by their different adaptations to snow. While lynx are physiologically adapted for hunting in deep snow (Murray and Boutin 1991, Pozzangera et al. 2016), bobcats are not (Marston 1942, Litvaitis et al. 1986, Reed et al. 2017). Lynx have a low foot-loading due to their lightweight body and large paws (Hoving et al. 2003). In contrast, bobcats have proportionately heavier bodies, shorter legs, and smaller feet, which give them double the foot-loading of lynx (Parker et al. 1983, Buskirk et al. 2000). Thus, perhaps snow levels partially limit bobcats from using high elevations in BC. However, bobcats were found to thrive in deep snow conditions in Montana (Newbury 2013), thus even though bobcats are not adapted for deep snow, it is unclear whether snow actually limits their distribution. Prey distributions also vary with elevation, as does presence of other mesocarnivores, and at present it is not clear what limits bobcat distribution either elevationally or latitudinally.  I think that the elevation pattern of lynx detections can be largely explained by their  20 resource requirements. Unlike bobcats, which are dietary and habitat generalists (Woolf and Hubert 1998, Hansen 2007), lynx are specialists that are spatially concordant with snowshoe hares (Lepus americanus), their primary prey (McKelvey et al. 2000, Mowat et al. 2000). Snowshoe hares occur in dense coniferous forests (Hodges 2000a, b), which do not occur at low elevations in southern BC (BC Ministry of Forests and Range 1998). Similarly, such habitats do not occur at extremely high elevations, which instead give rise to sub-alpine and alpine environments (BC Ministry of Forests and Range 1998).  Thus, within the broad zone of overlap between bobcats and lynx that I detected in southern BC, the two species may be separated due to different adaptations and resource requirements. However, my findings suggest that there may be a ‘sweet spot’ within this zone of sympatry, between 400-1,200 m, where snow conditions and prey availability support bobcats yet snowshoe hare populations support lynx, and thus where it seems most likely that bobcats and lynx may interact. Future research should attempt to estimate co-occurrence of bobcats and lynx at finer scales in the areas of range overlap identified here, and ascertain whether interspecific competition exists between these two congeneric species.  In addition to providing provincial distribution estimates of bobcats and lynx, my work shows the usefulness of various data sources for estimating mesocarnivore distributions. Trapping records comprised the largest data set and provided the greatest spatial coverage. However, with no measure of effort reported in the trapping records or compulsory records from hunting, harvest records could not be used to assess absence or abundance (Erickson 1982, DeVink et al. 2011, Dorendorf et al. 2016). Trapper surveys likely reflected distribution more directly than harvest records and were the only data source that reported absence. The greatest problem with voluntary surveys is their high number of non-respondents (Schmidt et al. 2015):  21 my response rate was low (23%), which hindered the spatial coverage of the province, but was in line with other efforts (McDonald and Harris 1999, Dorendorf et al 2016). The number of bobcats and lynx reported in the provincial vehicle-kill records was also low. Although species such as deer (Odocoileus spp.) are commonly killed by cars, vehicle collisions with bobcats and lynx are rare. However, despite their low frequency, vehicle-kill records have the advantage of being collected consistently throughout the entire province (Sielecki 2004, 2005). Still, the low number of records combined with the chances of misidentification by maintenance workers who are not trained in wildlife identification make vehicle-kill records the most unreliable data source that I examined, despite the fact that these data derive from actual carcasses.  Citizen science has become increasingly important in ecological research (Dickinson et al. 2012, Forrester et al. 2017, Pocock et al. 2017). For example, iNaturalist and United Kingdom Mammal Tracker allow citizen scientists to submit wildlife images for the purpose of assessing distributions of species of interest. In my study, images submitted by the public comprised my second largest data set, had the finest spatial and temporal resolutions, and next to trapping records, provided the most unique locations for both species. Although trapping records provided more data points, the distribution of image detections closely matched the distribution of trapping records. In the case of bobcats and lynx in BC, images solicited from the public reasonably estimated distributions even without combining those data with the other sources that I examined. I do caution that bobcats and lynx can be difficult to distinguish from images (Chapter 4), thus surveys for these species should use more methods than images alone. Species that are easier to tell apart may be more able to rely on citizen science efforts for detection.   However, I stress that citizen science reports must be accompanied with an image or physical evidence of the animal to be verifiable (Aubry and Jagger 2006, Roy et al. 2016).  22 Unverified sighting records are unreliable for rare or secretive species and often overestimate their distribution because of a bias towards misclassifying more common species as the rarer species (McKelvey et al. 2008, Garrote and Ayala 2015). For example, in response to my outreach for images, I received a few emails, but without images, of bobcats in northern BC, hundreds of kilometers from the nearest verifiable image or physical record; such reports are extremely unlikely to be accurate. Further, while I could independently classify the species in each image, I could not verify image locations and thus I relied on honest location details from people submitting the records. In my case, I have no reason to suspect any bias or that people mis-reported locations, but I could imagine sensitive cases involving endangered species or clear management implications where follow-up surveys would be required before relying upon location reports from citizen science.  I provide estimates of current bobcat and lynx distributions and areas of range overlap in BC. I recommend more monitoring at the northern edge of bobcats to see how they use these northern habitats, determine how often northward exploratory dispersal occurs, and investigate the factors that may limit their northern distribution. I show that bobcats and lynx broadly overlap in much of southern BC, but may be segregated by elevation. Future work should explore these areas of range overlap to determine whether the two species compete in this mountainous landscape. My work also has wider applications for other species and jurisdictions, because I demonstrate the usefulness of merging various record types for estimating mesocarnivore distributions. In the case of bobcats and lynx, all of the record types that I compiled displayed similar spatial patterns, but each data source contributed unique locations to the final distribution maps that I derived. I also show that the general public is eager to provide data for studies like mine. However, I strongly echo previous work that has urged caution about  23 public records that lack images or physical evidence that enables expert verification (Aubry and Jagger 2006, McKelvey et al. 2008, Roberts et al. 2009). With sufficient outreach and media coverage, similar citizen-science approaches could be used to help estimate the distributions of other mesocarnivores, especially non-trapped or hunted species where harvest records are not available.  24 Table 2.1. Details of the data sources used to assess bobcat (Lynx rufus) and lynx (L. canadensis) distributions in British Columbia. The number of records for each species indicates the total number of data points available from each data source during the period of my analysis. Trapping records and compulsory records were obtained from BC’s Ministry of Forests, Lands, and Natural Resource Operations. Vehicle-kill records were obtained from BC’s Ministry of Transportation and Infrastructure. Trapper surveys and images submitted by the public were collected in this study.   Trapping records Compulsory records Vehicle-kill records Trapper surveys Public images Bobcat records 1,233 264 52 61 805 Lynx records 6,462 222 15 125 807 Temporal extent 2008-2013 2008-2015 2008-2016 2008-2016 2008-2017 Temporal grain Year Day Day Year Day Spatial extent Open traplines Open Management Units Highways Traplines BC Spatial grain Trapline Management Unit Km Trapline Coordinates  25 Table 2.2. Number of Management Units containing records of bobcats (Lynx rufus) and lynx (L. canadensis) from each data source during 2008-2017 in British Columbia. Values indicate the total number of Management Units with records from each data source (total), and the number of Management Units with records from only that data source (unique), of a) bobcats, b) lynx, and c) both species.   Number of Management Units with records  Trapping records Compulsory records Vehicle-kill records Trapper surveys Public images Any data source a) Bobcat          Total 92 67 31 43 76 112    Unique 9 2 3 2 7  b) Lynx          Total 146 68 14 73 102 164    Unique 23 2 0 3 7  c) Both species          Total 63 31 8 28 39 85    Unique 16 2 2 3 8          26   Figure 2.1. Distribution of bobcat (Lynx rufus) a) trapping records (2008-2013), b) compulsory records from hunting (2008-2015), c) presence and absence reported from trapper surveys (2008-2016), d) vehicle-kill records (2008-2016), and e) images submitted by the public (2008-2017) in British Columbia. Bobcat sketch was courtesy of N. Reynolds at Into The Wild Artistry.  27   Figure 2.2. Distribution of lynx (Lynx canadensis) a) trapping records (2008-2013), b) compulsory records from hunting (2008-2015), c) presence and absence reported from trapper surveys (2008-2016), d) vehicle-kill records (2008-2016), and e) images submitted by the public (2008-2017) in British Columbia. Lynx sketch was courtesy of N. Reynolds at Into The Wild Artistry.  28   Figure 2.3. Elevations of bobcat (Lynx rufus; black bars) and lynx (L. canadensis; gray bars) images taken during 2008-2017 in British Columbia estimated from a) all images, and b) only images from Management Units that contained records of both species from ≥1 data sources (i.e., areas of range overlap between bobcats and lynx). Images were collected using citizen science, and elevations were estimated from location descriptions provided by the people who submitted the images. Location details varied in quality and were often vague, thus these values are coarse estimates. Bobcat and lynx sketches were courtesy of N. Reynolds at Into The Wild Artistry.  29   Figure 2.4. Bobcat (Lynx rufus) and lynx (L. canadensis) distributions in British Columbia during 2008-2017 estimated from trapping records, compulsory records from hunting, trapper surveys, vehicle-kill records, and images submitted by the public. The four shades from white to black represent Management Units that contained no records, records from one data source, two sources, and three to five sources. The number of data sources with records in a Management Unit does not reflect abundance of bobcats or lynx, and Management Units with no records may reflect genuine absence of the species or simply a lack of records. Bobcat and lynx sketches were courtesy of N. Reynolds at Into The Wild Artistry.  30   Figure 2.5. Range overlap between bobcats (Lynx rufus) and lynx (L. canadensis) in British Columbia during 2008-2017. I show Management Units that contained records of both species from one (light gray) and ≥2 data sources (dark gray). Provincial highways are also shown.         31 Chapter 3  Range shifts are not always consistent along range edges: bobcats and Canada lynx remain stable in British Columbia  Literature review and objectives  Geographic ranges are dynamic and are shifting for many species across the planet (Parmesan and Yohe 2003, Hickling et al. 2006). Range shifts at peripheries are often characterized latitudinally or elevationally as leading-edge expansion (i.e., poleward or upward) or trailing-edge contraction (i.e., low latitude or low elevation; Lenoir and Svenning 2013, 2015). Chen et al. (2011) reported that terrestrial species in the northern hemisphere have recently shifted their ranges northward at a median rate of 16.9 km per decade, and to higher elevations at a median rate of 11.0 m per decade. However, range shifts vary within a species; in many cases species shift in parts of their geographic range, but range edges remain stable or even shift in opposite directions in other parts of their range, especially for wide-ranging species (Gibson-Reinemer and Rahel 2015).  Many biotic and abiotic changes can lead to range shifts, and these causes can vary across the geographic range of a species. Common mechanisms include changes in biological use (e.g., overharvest can cause extirpation, while closures can allow a species to recover), land conversion such as forest harvesting that alters habitat structure and ecological communities, as well as climate change. During the 2000s, ~7.5 million hectares of forest across North America were affected by anthropogenic disturbances (e.g., clear cutting, land conversion) each year (Masek et al. 2011). At the same time, climate change has caused an increase in average global  32 temperature (Meehl et al. 2007), leading to earlier springs (Cayan et al. 2001), lower snow levels in many areas (Mote et al. 2005, Knowles et al. 2006), and changes in snow quality across North America. In addition, climate change is also shifting forests (Flower et al. 2013, Gray and Hamann 2013, Jackson et al. 2016) and increasing natural disturbances such as wildfire (Cannon and DeGraff 2009), which is reshaping the boreal forest.   Any of these changes may induce range shifts for many species, including bobcats (Lynx rufus) and Canada lynx (L. canadensis; hereafter lynx). Bobcats and lynx are congeneric mesocarnivores that are wide-spread across North America, but their ecology is very different. While lynx are adapted for deep snow (Parker et al. 1983, Murray and Boutin 1991, Pozzangera et al. 2016), bobcats are not (Marston 1942, Litvaitis et al. 1986, Reed et al. 2017). Bobcats are habitat and dietary generalists; they occur across most of the contiguous United States, Mexico, and southern Canada (Woolf and Hubert 1998, Anderson and Lovallo 2003, Hansen 2007). In contrast, lynx are specialists, and their range follows that of snowshoe hares (Lepus americanus), their primary prey (McKelvey et al. 2000, Mowat et al. 2000, Koehler et al. 2008). Snowshoe hares, and thus lynx, are confined to coniferous forests across Canada, Alaska, and the northern US or in high elevation forests along the mountain chains (Hodges 2000a, b).  Recent analyses have shown that bobcat populations are increasing across North America, and their range is expanding in eastern parts of the continent. Since the early 1980s, bobcat populations have increased in 32 US states, including eight of the 11 states bordering Canada, and the country’s bobcat population has roughly doubled during that time (Roberts and Crimmins 2010). In Quebec, harvest records indicated that bobcats have expanded northward ~115 km since the 1980s (Lavoie et al. 2009). Expansion of bobcats at the continental scale has been suggested to be the result of a warming climate (Hatler et al. 2003, Roberts and Crimmins  33 2010), but the northward shift in Quebec is attributed to the closure of bobcat trapping (Lavoie et al. 2009).   In contrast, lynx are contracting their range across eastern North America. Lynx have lost some of their southern range within the contiguous US (McKelvey et al. 2000, Laliberte and Ripple 2004) and are federally listed as Threatened (US Fish and Wildlife Service 2000). The majority of their remaining range in the contiguous US lies along the Canadian border (Interagency Lynx Biology Team 2013, Lewis 2016). In southern Ontario, lynx have also contracted their range northward >175 km since the 1970s (Koen et al. 2014). Historical range contractions of lynx in the US were likely mostly due to over-harvesting and loss of habitat from forest harvest (US Fish and Wildlife Service 2000), but the contraction in Ontario is at least partially attributed to changes in snow depth (Koen et al. 2014). There is speculation that bobcats may outcompete lynx (Parker et al. 1983, Buskirk et al. 2000, Peers et al. 2013), leading to the retraction of lynx in areas of sympatry with bobcats (Hoving et al. 2003, Poole 2003), but more data are needed to conclude whether such interspecific competition occurs and if so, whether it drives geographic ranges.  Bobcat and lynx ranges overlap in British Columbia, Canada (Hatler and Beal 2003, Hatler et al. 2003), but it is unclear whether either species has had a recent range shift in this province. Because climate change and habitat alteration are obvious in BC (Murdock et al. 2013, Najafi et al. 2017), and given the range shifts for both species elsewhere, it is reasonable to predict range shifts for both species in BC. I therefore examine whether 1) bobcats have expanded their range, 2) lynx have contracted their range, and 3) range overlap of these congeneric species has changed over the past century.   34 Methods  My analysis focused on spatially explicit trapping records, supplemented by surveys of trappers. Bobcats and lynx are managed as furbearers on a registered trapline system in BC (BC Ministry of Forests, Lands, and Natural Resource Operations 2017; see Appendix A for the harvest regimes of bobcats and lynx). BC currently contains 2,451 traplines ranging from 0.6 km² to 23,310 km² (   = 366 km²), within 225 Wildlife Management Units ranging from 465 km² to 18,982 km² (   = 4,216 km²; see Appendix B for maps of BC’s Management Units and registered traplines). Each trapline is registered to a trapper who has the exclusive rights to harvest furbearing animals on Crown land within that area. Trapping seasons remained relatively consistent throughout the 78 years of records that I examined.  I obtained trapping records and spatial data for traplines and Management Units from BC’s Ministry of Forests, Lands, and Natural Resource Operations. Trapping records contain royalty reports from all animals harvested on traplines, reported by trapline, as well as on private land, reported by Management Unit. Trapping records from traplines and private land were available provincially from 1983 to 2013, and from traplines in the Thompson and Cariboo Regions in central BC from 1935 to 2013. Other Regions lack digitized trapping records prior to 1983. I omitted 37 anomalous records (i.e., 32 bobcats from Vancouver Island and northern BC, and five lynx from Vancouver Island and the Lower Mainland) that I assumed were errors (Chapter 2).  I inspected the trapline files at each regional Ministry of Forests, Lands, and Natural Resource Operations office to obtain a history of trapline amalgamations. Amalgamation occurred when an owner of two adjacent traplines was granted permission to join them into one trapline. Approximately 13% of traplines reported in the trapping records had amalgamated since  35 the registered trapline system was established in 1926. Traplines were also occasionally cancelled and relinquished to the crown; I found 36 such cases and omitted these traplines. I combined trapping data from amalgamated traplines to report harvest from traplines as they existed in 2013.  I compiled harvest by Management Unit by identifying the Management Unit that contained the majority of each trapline. Summarizing harvest within Management Units allowed me to combine harvest from traplines and private land, provided a clear spatial resolution at the provincial scale, and reduced the impacts of fluctuations in trapper effort across time periods; most traplines were not active year-after-year, but because there were multiple traplines within each Management Unit, most Management Units had active trapping across time periods. I summarized harvest from traplines and private land provincially during three decadal periods between 1983 and 2013, and from traplines in central BC during three quarter-century periods between 1935 and 2013. I analyzed time periods ≥10 years to ensure that lynx distribution records were not distorted by the phase of their 10-year cycle.  I defined Management Units as being ‘active’ in each time period if they contained harvests of ≥1 bobcat, lynx, red fox (Vulpes vulpes), coyote (Canis latrans), wolf (C. lupus), or wolverine (Gulo gulo). I assumed that trappers who harvested these species used traps that were able to catch bobcats and lynx. I defined Management Units as being ‘inactive’ if they had no harvest or harvested only species not listed above, such as marten (Martes americana or M. caurina) or beaver (Castor canadensis), because traps for these species do not have the potential to catch bobcats or lynx.  I also surveyed trappers across BC to assess whether they have noticed any evidence of range shifts. In fall 2016, I mailed surveys to owners of traplines that harvested any species in ≥5  36 years during 2000-2013, and harvested ≥1 bobcat, lynx, red fox, coyote, wolf, or wolverine during that time (see Appendix C for the survey). My survey obtained ethics approval from the University of British Columbia (certificate # H16-02432). I obtained contact information for trappers from BC’s Ministry of Forests, Lands, and Natural Resource Operations. I provided the option of mailing back surveys or completing an online version. To encourage participation, in 2016 I presented at BC Trapper’s Association meetings and published two articles in the BC Trapper magazine.  I asked trappers for the year in which they first started trapping their trapline. I considered only traplines that had been trapped by the current trapper since 2011 or earlier, to ensure that trappers in my analysis had been trapping long enough (≥5 years) to provide reliable information about changes through time. I asked trappers i) to indicate whether bobcats and lynx were present on their trapline each year during 2000-2016, ii) if bobcat and lynx populations on their trapline had increased, decreased, or stayed the same since they first started trapping, iii) if bobcats and lynx on their trapline now occur at higher, lower, or at the same elevations as when they first started trapping, and iv) if lynx on their traplines go through population cycles.  Statistical analysis  I calculated coincidence summaries for each consecutive time period, as well as for the first and last time periods for the provincial and central BC trapping records. The coincidence summary statistic measures the similarity between two maps (Berry 1993), defined here as the percentage of Management Units that had the same value (harvest or no harvest) in both time periods under comparison. Only Management Units that were active in both time periods were included in my calculations. A higher coincidence summary indicates greater alignment, with  37 lower coincidence summaries indicating more Management Units that switched from the species being harvested to not harvested, or vice-versa.   I used the Pearson correlation coefficient to determine whether bobcat and lynx harvests were correlated with the current pelt price, previous year’s pelt price, or total harvest of bobcats, lynx, or marten. Marten are valuable furbearers that can influence overall trapper effort (Webb et al. 2008). I used harvest totals and average pelt prices for each species, derived from their total number of pelts and total annual pelt values, in BC from 1983 to 2009 from Statistics Canada (2012). I adjusted pelt prices for inflation to reflect 2009 values, the latest year that had summary pelt prices. I used a nominal α-level of 0.05, but separately for lynx and bobcats I used Bonferroni corrections for multiple tests, resulting in a threshold of α’ = 0.006.  Results  Bobcats do not appear to have expanded their range northward in BC (Figure 3.1, Figure 3.2, Table 3.1). During all 3 decadal periods of 1983-2013, bobcats were harvested across the southern half of BC, extending from the western coast of the province east to the Alberta border, and north from the US border to approximately Quesnel in the Cariboo Region (~53.0° N; see Appendix G for annual distribution maps of bobcat harvest). Older trapping records from central BC indicated that bobcats have occurred in the northern Cariboo since at least 1935, thus at the scale of Management Units, their northern range limit has not changed over this 80 year span.  Similarly, lynx do not appear to have contracted their range (Figure 3.1, Figure 3.2, Table 3.1). During 1983-2013, lynx were harvested throughout most of the interior of the province, extending from the Coast Mountains east to the Alberta border, and north from the US to the Yukon border (see Appendix G for annual distribution maps of lynx harvest). Lynx  38 occurred across southern BC along the US border except for the southern coast and a portion of southeastern BC. Of the 20 Management Units that bordered the US, 13 reported lynx harvest during 1983-1993, while 11 reported lynx harvest during each of 1994-2003 and 2004-2013; those 11 Management Units were not all the same in both periods. Between 2004 and 2013, 380 lynx were trapped ≤100 km from the US border.  The range overlap between bobcats and lynx also appeared to remain stable (Table 3.1). Of the provincial traplines that harvested either bobcats or lynx during 1983-1993, 1994-2003, and 2004-2013, 36%, 32%, and 35% harvested both species in each time period. Management Units in which both species were harvested showed a similar spatial distribution throughout these 3 decadal periods.   The trapper surveys confirmed that there have been no recent shifts in the ranges of either species. During 2000-2013, 796 (33%) traplines were active ≥5 years and harvested ≥1 bobcat, lynx, red fox, coyote, wolf, or wolverine. I received 157 completed surveys of the 675 that I mailed (23% response rate), 138 of which were from trappers who had been trapping since 2011 or earlier (see Appendix E for a map showing active traplines and those that responded to my survey). Most trappers reported no changes in the range of either species on their traplines, but many trappers did report changes in the abundance of each species (Table 3.2). However, traplines with a reported increase or decrease of each species were spread throughout the province and showed no clear patterns. Of the 100 trappers who responded to the question about lynx population cycles, most (75%) reported that lynx populations cycled on their trapline, but some (25%) throughout the province reported that lynx populations did not cycle.  Annual harvest and pelt prices fluctuated greatly during 1983-2013 for each species (Figure 3.3). During this time period, 67-309 bobcats (   = 140.4 ± 11.51), and 382-2,025 lynx (    39 = 957.0 ± 63.13) were trapped annually across BC. Average pelt prices per year varied from $48-692 for bobcats (   = $180.5 ± $26.60), and $79-1,257 for lynx (   = $296.9 ± $64.35). Bobcat harvest was correlated with all predictors tested at the nominal α-level of 0.05, but with only the current year’s pelt prices of bobcat, lynx and marten, and the previous year’s pelt prices of bobcat and lynx at the Bonferroni-corrected α-level (Table 3.3). The strongest correlation with bobcat harvest was their previous year’s pelt price (Pearson’s r = 0.71). Lynx harvest was correlated with the current pelt prices of lynx and marten, the previous year’s pelt prices of lynx and bobcat, and total bobcat and marten harvest at the nominal α-level of 0.05, but with only  total marten harvest at the Bonferroni-corrected α-level (Pearson’s r = 0.51; Table 3.3).  Discussion  Bobcat and lynx ranges appeared to have remained stable in BC over the last century. Trapping records indicated that the provincial range of each species has not changed since at least 1983, and their ranges in central BC have not changed since at least 1935. Surveys from long-time trappers supported these findings. Lynx appeared to have not contracted from southern BC, but rather persist along the US border, which is noteworthy because dispersal from BC potentially supports lynx populations in Washington (US Fish and Wildlife Service 2005, Interagency Lynx Biology Team 2013, Vanbianchi et al. 2017). Changes in abundance at northern or southern range peripheries or finer-scale changes within the provincial range may have occurred for either species, but range shifts at the scale of Management Units have not occurred. I was not able to evaluate fine-scale changes given the resolution of my data; trapping records can detect only broad-scale changes in distribution (Litvaitis et al. 2006, Golden et al. 2007, Robichaud and Boyce 2010).  40  While many trappers suggested that bobcat and lynx populations have changed throughout the province, I did not attempt to measure abundance because with no index of trapper effort, harvest records cannot reliably detect population changes (Poole and Mowat 2001, DeVink et al. 2011, Dorendorf et al. 2016). Harvest totals are correlated with environmental as well as socioeconomic factors such as pelt prices (Brand and Keith 1979, Tumlison and McDaniel 1986, Gerht et al. 2002), season length (Hiller et al. 2011, Kapfer and Potts 2012), number of trapping licenses sold (Lewis and Zielinski 1996), and harvest totals of other species (Chilelli et al. 1996). My results confirmed that pelt prices and total pelts were correlated with harvest totals of bobcats and lynx.  Provincially, most mainland Management Units were active in each of the three decadal periods since 1983 (median = 91%, range = 90-96%); this level of activity offered high spatial and temporal coverage of the province. However, few (n = 39) Management Units contained data prior to 1983, which resulted in lower coincidence summaries for the central BC records because coincidence summaries from fewer locations were more affected by small changes. I considered Management Units to be active if they contained records of harvested mesocarnivores, because trappers in those areas used traps with the potential to catch bobcats and lynx. While not a true measure of absence, areas with many active trappers but with no bobcat or lynx harvest strongly suggested that bobcats or lynx were rare or did not occur in those areas, a result that coincides with other recent records of both species (Chapter 2).  There has never been a trapping season for bobcats north of the Cariboo Region, and my documented range limit from 1935 to 2013 matched the northern limit of bobcat trapping during that time. It could therefore be argued that bobcats north of the Cariboo would be undetectable from harvest records. However, there is strong evidence from other data sources that the range  41 limit for bobcats genuinely occurs here (Chapter 2).  The lack of range shifts that I detected for bobcats and lynx in BC differs from patterns observed in eastern North America. BC contains heterogeneous landscapes and mountainous topography, thus BC is a much different system than the flatter and much lower-elevation terrain in eastern North America where range shifts have been documented (Lavoie et al. 2009, Koen et al. 2014). Bobcats and lynx are under different environmental pressures and interact with different ecological communities across their ranges, thus their range limits are likely not controlled by the same factors along their entire range edges.  The factors that limit bobcat and lynx ranges in BC are not clear, and likely differ from those influencing ranges in eastern North America. Bobcats and lynx do not occur throughout the entire province, but are highly mobile species capable of long-distance dispersal and could easily reach anywhere in the mainland (Mowat et al. 2000, Johnson et al. 2010). Thus, the range of each species must be limited by one or more factors such as prey availability, type of forest cover, snow depth and compactness, winter duration, or interspecific competition. My data did not permit investigation of factors that control bobcat and lynx ranges, but I think that the absence of a northern range expansion for bobcats suggests that winter conditions or other climate-related changes may not be entirely explanatory in this province.  Perhaps the most common prediction for what controls bobcat and lynx ranges and why they have shifted in some areas is snow conditions (Litvaitis et al. 1986, Hoving et al. 2003, Poole 2003, US Fish and Wildlife Service 2005, Koen et al. 2014, Reed et al. 2017). Even if snow depths, snow compactness, or winter duration partially limits either species in BC, it is unclear whether these metrics have changed significantly here. Models suggested that snow depths in BC have dropped over the past century (Mote et al. 2005, Choi et al. 2010, Kang et al.  42 2014, Najafi et al. 2017), but climate projections in mountainous regions are complex and snow depth measurements are exceedingly rare, resulting in substantial uncertainty about the magnitude and geography of snow loss across BC (Najafi et al. 2017). Mountains harbor many more microclimates than flatter regions, and climate velocities are slower in mountains as a result (Loarie et al. 2009), suggesting that range shifts in mountainous terrain may be slower than in flatter regions.  If range limits of bobcats or lynx are partially controlled by snow, perhaps snow depths in BC have not dropped below the threshold for enabling bobcats to expand northward, or forcing lynx to contract from southern BC. Another possibility is that snow does not limit the ranges of these species in BC. However, climate change is accelerating; in the Fraser River Basin of BC, snowfall is projected to decrease by ~50%, and spring snow melt is projected to be ~25 days earlier in the 2050’s (Islam et al. 2017). It would be interesting to re-evaluate the ranges of bobcats and lynx in BC in the coming decades to see if their ranges shift in the future. It would also be valuable to investigate the energetics of both species in winter, to determine whether there are particular combinations of snow conditions and prey availability that negatively affect energy balance, potentially leading to high mortality or reduced reproductive success.    My results highlight that range shifts for wide-ranging species may be very localized and small-scale rather than occurring across an entire range edge. I propose that variable, not uniform, range edges may be the norm rather than the exception for wide-ranging species. Consistent with my findings, Gibson-Reinemer and Rahel (2015) reported that range shifts were not consistent across the geographic ranges of 42-50% of the 273 species that they examined. I show that in the case of bobcats and lynx, there may be substantial differences in their range dynamics in eastern and western populations.  43 Table 3.1. Coincidence summaries for the distributions of bobcats (Lynx rufus) and lynx (L. canadensis) summarized by Management Units in British Columbia at two scales: provincial (1983-2013) and central BC (1935-2013). Values for bobcat and lynx indicate the percentage of active Management Units that had the same value (harvest or no harvest) in both time periods. Values for overlap indicate the percentage of active Management Units that either harvested both species or did not harvest both species in both time periods. The number of active Management Units is those that were active in both time periods under comparison.  Time period comparison Number of active Management Units Coincidence summary (%)   Bobcat Lynx Overlap Provincial        1983-1993 vs 1994-2003 194 92.3 90.2 91.8    1994-2003 vs 2004-2013 186 97.8 96.2 95.2    1983-1993 vs 2004-2013 191 90.1 88.0 92.1 Central BC        1935-1961 vs 1962-1987 38 86.8 89.5 81.6    1962-1987 vs 1988-2013 43 79.1 93.0 79.1    1935-1961 vs 1988-2013 38 71.1 86.8 65.8      44 Table 3.2. Changes noticed by trappers (n = 138) in the presence, elevation, and abundance of bobcats (Lynx rufus) and lynx (L. canadensis) on their traplines in British Columbia. Changes in presence are relative to when each trapper first started trapping or the year 2000 (i.e., whichever is most recent), while changes in elevation and abundance are relative to when each trapper first started trapping from any time. Values are the number of trappers that reported each case on my survey.    Bobcat Lynx Presence     Stable 37 105  Expandeda 9 5  Contractedb 2 1 Elevation     Stable 24 85  Higher 9 11  Lower 9 7 Abundance     Stable 22 47  Increased 22 33  Decreased 5 23  aCases where trappers indicated that a species was newly present on their trapline. bCases where trappers indicated that a species ceased to occur on their trapline. 45 Table 3.3. Correlates of bobcat (Lynx rufus) and lynx (L. canadensis) trapper harvest during 1983-2009 in British Columbia. I used harvest totals and average pelt prices for each species in BC from Statistics Canada (2012), and adjusted pelt prices for inflation to reflect 2009 values, the latest year that had summary pelt prices.   Bobcat harvest Lynx harvest  Pearson’s r P Pearson’s r P Current pelt price        Bobcat 0.69 <0.001 0.35 0.078    Lynx 0.64 <0.001 0.47 0.014    Marten 0.61 <0.001 0.39 0.046 Previous year’s pelt price        Bobcat 0.71 <0.001 0.44 0.022    Lynx 0.59 0.001 0.38 0.050    Marten 0.48 0.011 0.26 0.182 Total harvest        Bobcat   0.47 0.014    Lynx 0.47 0.014      Marten 0.47 0.014 0.51 0.006              46   Figure 3.1. Distributions of bobcats (Lynx rufus) and lynx (L. canadensis) from trapping records summarized by Management Units in British Columbia in a) 1983-1993, b) 1994-2003, and c) 2004-2013. Black Management Units harvested ≥1 bobcat (top) or ≥1 lynx (bottom) during each time period. Gray Management Units contained active trappers, but had no harvest of bobcats (top) or lynx (bottom). White areas contained no active trappers. Bobcat and lynx sketches were courtesy of N. Reynolds at Into The Wild Artistry.        47   Figure 3.2. Distributions of bobcats (Lynx rufus) and lynx (L. canadensis) from trapping records summarized by Management Units in central British Columbia in a) 1935-1961, b) 1962-1987, and c) 1988-2013. Black Management Units harvested ≥1 bobcat (top) or ≥1 lynx (bottom) during each time period. Gray Management Units contained active trappers, but had no harvest of bobcats (top) or lynx (bottom). White areas indicate no data. Bobcat and lynx sketches were courtesy of N. Reynolds at Into The Wild Artistry.       48   Figure 3.3. Total harvests (bars and left y-axis) and average pelt prices (line and right y-axis) of bobcats (Lynx rufus; top) and lynx (L. canadensis; bottom) from 1983 to 2013 in British Columbia. Harvest totals were obtained by BC’s Ministry of Forests, Lands, and Natural Resource Operations. Pelt prices were derived from total pelts and average pelt values in BC from Statistics Canada (2012), and adjusted for inflation to reflect 2009 values, the latest year that had summary pelt prices. Bobcat and lynx sketches were courtesy of N. Reynolds at Into The Wild Artistry.      49 Chapter 4  Poor agreement among experts in classifying camera images of similar species  Literature review and objectives  Ecological research is experiencing an explosion in the use of wildlife imagery. Camera trapping has become a common non-invasive survey technique (Rowcliffe and Carbone 2008, O’Connell et al. 2011, Burton et al. 2015), especially for rare and elusive forest-dwelling species (Stewart et al. 2016, Furnas et al. 2017), and has been used to obtain crucial ecological information (Caravaggi et al. 2017). Landscape-scale camera grids or transects are increasing across the globe (McShea et al. 2016), and such sampling may be used to monitor global biodiversity in the future (Rich et al. 2016, Steenweg et al. 2017). Similarly, numerous websites and mobile phone applications have been developed for people to submit wildlife images for the purpose of assessing species’ distributions. For example, the United Kingdom Mammal Tracker application allows the general public to submit geo-located images of 39 wildlife species.  Such citizen-science projects and camera networks can collect seemingly endless data across broad scales, but the data are often of limited utility because of the need to classify the animals that the images contain (He et al. 2016, Wearn and Glover-Kapfer 2017). In almost all cases, researchers are interested in classifying each animal to the species level, and in many cases even to individuals (Weingarth et al. 2012, Rich et al. 2014). However, classifying images is difficult when they are blurry, taken in poor lighting, show only part of the animal, or are the only image of that individual. Further, even high-quality images may be difficult to classify if the species has similar sympatrics (Yu et al. 2013, McShea et al. 2016, Swanson et al. 2016).  50 Classifiers may also have a bias towards one sympatric species over another. For example, rare species can have higher false-positive and false-negative errors than common species (McKelvey et al. 2008, Swanson et al. 2016). Correct species classification is crucial; even low misclassification rates can lead to significant over- or underestimation of the occupancy, habitat preferences, or distribution of a species (Royle and Link 2006, Miller et al. 2011, Molinari-Jobin et al. 2012, Costa et al. 2015), which could hinder conservation efforts (McKelvey et al. 2008).  Studies have classified images using various methods including the researchers’ own judgments (Liu et al. 2017), crowd sourcing from the general public (Swanson et al. 2016), and automated classification by computer software (Hiby et al. 2009, Jiang et al. 2015). In most cases, experts classify all images or at least a subset of images (McShea et al. 2016). Despite the fact that even highly trained experts are not always correct (Gibbon et al. 2015, Austen et al. 2016, Swanson et al. 2016), expert classification is often considered conclusive and is rarely questioned.  Classification of images by a single expert may be adequate when classifying species that are distinctive, such as mountain goats (Oreamnos americanus), porcupines (Erethizon dorsatum), and snow leopards (Panthera uncia), but may be unreliable for sympatric species that are similar in size, shape, or colouration. Many species across the globe fall into this category such as bears, deer, lemurs, some mustelids, felids and antelopes, as well as many bats, raptors, and owls. Specific examples include grizzly bear (Ursus arctos) versus black bear (U. americanus), mule deer (Odocoileus hemionus) versus white-tailed deer (O. virginianus), nyala (Tragelaphus angasii) versus greater kudu (T. strepsiceros), and sharp-shinned hawk (Accipiter striatus) versus Cooper’s hawk (A. cooperii).  Here, I use bobcats (Lynx rufus) and Canada lynx (L. canadensis; hereafter lynx) as a  51 case study to measure agreement among experts in their classifications of images showing similar-looking species. Bobcats and lynx are congeneric felids similar in size and appearance that are sympatric across southern Canada and the northern United States (McKelvey et al. 2000, Hansen 2007, Chapter 2). Bobcats are common and are legally harvested in both countries, but lynx are federally listed as Threatened in the contiguous US (US Fish and Wildlife Service 2000). Classification of felid images in the contiguous US thus has direct conservation implications for lynx; bobcats falsely classified as lynx could result in false models of occupancy or protection of areas that are not in fact used by lynx, whereas lynx misclassified as bobcats could mean under-protection.  Methods  I measured agreement among experts in their classifications of bobcat and lynx images that I collected though citizen science. In a separate study, I solicited 4,399 images of bobcats and lynx from the public across British Columbia, Canada to examine the provincial distribution of each species (Figure 4.1; Chapter 2). From those images, I selected 299 that I subdivided into 15 categories based on different image characteristics (Table 4.1). I grouped those categories into six trials of images that I sent to 27 experts. I asked the experts to classify those images, and then I compared the agreement in their classifications between the different categories of images. I also compared agreement among experts from different locations in their classifications of the images. The sixth trial repeated the first trial to test whether experts were consistent in their classifications of the same images.  I selected 27 experts from across western North America to classify the images; I chose experts from i) northern BC and the Yukon (n = 9), where lynx are common but bobcats are  52 likely abset, ii) southern BC (n = 8), where both species are common, and iii) the northwestern US (n = 10), where lynx are rare but bobcats are common (Figure 4.1). I did not reveal the identity of the experts to each other. I considered people as bobcat or lynx experts if they were biologists who had experience working with or classifying images of either species. My panel of experts represented people likely to participate in camera-trapping studies of one or both species, or who would likely be asked to classify bobcat or lynx images. The experts consisted of mesocarnivore and furbearer biologists from provincial, state, and federal government agencies, as well as private consultants and academics.  To select images for the different categories (e.g., ‘summer’ category in the ‘season’ trial), I first chose images from the entire set that were of good photographic quality (i.e., the animal was in focus and not distant), were of single, alive, adult individuals that showed no bait or prey, and that were not submitted by any participating experts. I did not crop, edit or modify the images in any way. I then constructed a pool of images for each category, from which I randomly selected images to populate each category (Table 4.1). Within each category, all image characteristics (i.e., season, background habitat, visible features, and time of day) were consistent.  Each image was used only once, except for images in the ‘season’ trial which were repeated as the ‘consistency’ trial. I also mistakenly included one image twice in the ‘legs and tail’ category. I disregarded the second classifications from the experts for this image in all analyses, which resulted in the ‘legs and tail’ category containing 19 images rather than 20. Multiple images taken by the same remote camera, and thus that had the same background, were not included in the same trial. If the ratio of what I thought were bobcat and lynx images were below 4:1 for either species in any category, I randomly replaced images until at least that ratio  53 was achieved, except for the ‘northern’ images in the ‘location’ trial because bobcats rarely occurred in northern BC (Figure 4.1; Chapter 2).  Because I used images that were contributed by the public, I was unable to independently verify the species in each image. Instead, I classified the species in the images as follows. (A) I used the location of the image. Separately, I examined harvest and other records of bobcats and lynx in BC (Chapter 2), and those results showed distinct distributions for each species. Thus in many cases the location of an image strongly supported the classification of the animal (e.g., lynx in the far north, and bobcats on the southern coast). (B) I usually examined multiple images of the same individual animal; submissions contained a median of two images per sequence (range = 1-38). Because most of the images that I asked experts to classify were derived from sequences of multiple images of the same animal, I had the advantage of being able to use several images in my own classifications, and often the images showed different parts of each animal. (C) I used only images for which my supervisor, Karen Hodges, and I independently agreed on the species in the image. Although I am deeply aware that such agreement is not definitive, this filter at least meant that I was not choosing images that I myself found difficult to classify. My classifications were thus based on more information than I made available to the experts.  I created weblinks for the six trials (Table 4.1) using FluidSurveys (www.fluidsurveys.com). I released trials online sequentially, two weeks apart, between January and April 2017. In each trial, experts were prompted to classify the species in each image, one at a time, by selecting ‘bobcat’, ‘lynx’, or ‘unknown’; the experts were not able to zoom in on images (see Appendix I for a screenshot of one of the trials). The order of images in each trial was random, but was the same for all experts. Experts could not proceed to the next image  54 without selecting an answer, and once selected, experts could not view previous images. However, experts were allowed to save unfinished trials and complete them at a later time. Trials were password protected, and I instructed experts to not consult with others. My experiment obtained ethics approval from the University of British Columbia (certificate # H16-03169).  Experts were aware that I was measuring agreement among them in their classifications of images, but they were unaware of the conditions that I was testing in each trial. Experts were unaware of image locations, except for half of the images in the location trial (Table 4.1). These images were accompanied with a map of BC showing the location of the image with a red star. The map also included cities and highways to help orient the experts.  Statistical analysis  I could not determine whether expert classifications were accurate because the species in each image was not independently confirmed. Instead, I measured agreement among experts in their classifications of the images (hereafter agreement) using Fleiss’ Kappa (K), which measures reliability among a group of classifiers (Fleiss 1971). A value of 1 indicates perfect agreement, while a value of 0 indicates agreement that would occur by chance.  I determined whether agreement varied between images with different characteristics (i.e., season, background habitat, visible features, and time of day) by comparing K between categories of images within each trial (Table 4.1). I also determined the combination of image characteristics that resulted in the highest and lowest agreement by pooling images with the same combination of characteristics from all categories.  I determined whether knowing the location of an image affected agreement by comparing K when experts knew the location of an image to when they did not for images taken in northern  55 and southern BC (Figure 4.1).  I determined whether experts were consistent in their classifications by having them unknowingly reclassify images from the first trial (‘season’ trial) 10 weeks later, and calculating K between their first and second classifications of the same images.  I determined whether agreement varied depending on the rarity of a species where experts lived by comparing K between experts from the three regions. I also determined whether knowing the location of an image affected agreement within expert groups differently for either northern or southern images, and compared the overall consistency within the three expert groups.  I also calculated the proportion of agreement for each image using the following equation, where bobcat, lynx, and unknown, are the number of experts who classified an image as ‘bobcat’, ‘lynx’, and ‘unknown’, respectively, and n is the total number of experts:                                     With three classification options, the proportion of agreement had an upper bound of 1.00, indicating perfect agreement, and had a lower bound of 0.31, indicating perfect disagreement (i.e., of 27 experts, nine each classified an image as ‘bobcat’, ‘lynx’, and ‘unknown’).  Finally, I calculated the number of experts required to classify an image to reach a final classification (i.e., the number of experts at which the majority classification would be unlikely change by asking more experts). I calculated the average probability that the majority classification (i.e., the classification of the greatest number of experts) of a randomly selected subset of one to 27 experts matched the majority classification of all 27 experts.    56 Results  All 27 experts completed each of the six trials. The following results refer to all images in the first five trials (n = 259 images); this set excludes the 40 images in the ‘location’ trial for which I provided locations. The total number of individual expert classifications was 6,993 (27 experts x 259 images); the experts classified the images as ‘unknown’ in 11% (n = 753) of classifications, and as ‘bobcat’ or ‘lynx’ in 89% (n = 6,240) of classifications.  Of those 259 images, 71% (n = 185) had ≥1 experts classify that image as ‘unknown’. Experts reached a majority classification of ‘unknown’ for 3% (n = 9) of images, but experts did not unanimously classify any images as ‘unknown’. Experts unanimously classified 24% (n = 61) of images as being either ‘bobcat’ or ‘lynx’, while 39% (n = 101) of images had ≥1 experts classify that image as ‘bobcat’ and ≥1 as ‘lynx’.  Overall the 27 experts had low agreement in their classifications of the 259 images (K = 0.64, 95% CI = 0.63-0.64). Experts did not unanimously classify the majority of images (76%; Figure 4.2a); the average proportion of agreement score for individual images was 0.79 (sd = 0.19), but was highly variable (Figure 4.2b, Table 4.2). However, experts had similar agreement for each species; the average proportion of agreement score was 0.84 (sd = 0.18, n = 92) and 0.77 (sd = 0.19, n = 167) for images that I had classified as ‘bobcat’ and ‘lynx’, respectively.   Experts had varying levels of agreement between images with different characteristics (Table 4.3, Figure 4.3). Experts had far greater agreement for winter images than summer images. Experts had greater agreement for images with a developed background (i.e., showing human infrastructure) or grassland background than images with a forest background. Experts had greater agreement for images showing the full body or only the face and legs of an animal than images showing only the face or only the legs and tail of an animal. Experts had greater  57 agreement for images taken at night than images taken during the day. Experts had the lowest agreement for daytime summer images with a forest background showing only the legs and tail of an animal (K = 0.34, 95% CI = 0.32-0.36, n = 15 images; Table 4.3). Experts had the greatest agreement for daytime winter images with a forest background showing only the face and legs of an animal (K = 0.80, 95% CI = 0.78-0.82, n = 35 images).  Experts had greater agreement when the location of an image was provided, and experts had greater agreement for southern images than northern images (Table 4.3). As an example of possible location bias, the left image in Figure 4.4a was taken near Prince George in 2016, and sent to me as part of my citizen science search for images (Chapter 2). At the time there had never been a confirmed bobcat sighting that far north. I classified the image as ‘bobcat’, but the image was widely circulated on social media and the local news station, which sparked an intense debate among hunters, trappers, and naturalists as to whether the animal was a bobcat or lynx. Biologists in Prince George were asked by the local news station to classify the image, and initially four biologists thought ‘bobcat’, and four biologists thought ‘lynx’. After additional images showing the animal’s feet were shared, those biologists shifted towards classifying the image as ‘bobcat’ or ‘possible hybrid’ (K. Otter, personal communication). The right image in Figure 4.4a is of the same animal and shares the same characteristics (i.e., season, background habitat, visible features, and time) as the left image. I asked experts to classify the right image in my experiment without providing its location; 26 experts classified the image as ‘bobcat’, and one as ‘unknown’.   Further, experts were inconsistent even with themselves, as shown by comparing the expert’s classifications of the 40 images in the ‘season’ trial with their classifications of those same images 10 weeks later. Experts had an average consistency with themselves of K = 0.67  58 (95% CI = 0.61-0.72). On average, experts changed their classifications on seven of the 40 images (sd = 3.4, range = 1-15). Further, I mistakenly included one image twice in the ‘legs and tail’ category, and three experts changed their classification of this repeated image within the same trial. However, experts showed improved agreement across the course of the six trials; whereas experts had an agreement of K = 0.55 (95% CI = 0.53-0.56) for images in the first trial (‘season’ trial), experts had an agreement of K = 0.63 (95% CI = 0.62-0.65) for the same images 10 weeks later (‘consistency’ trial).  Experts had contradictory majority classifications for different images of the same animal in two cases (Figure 4.4b). Out of all 299 images, there were 27 sets of images in the six trials where this discrepancy could happen (i.e., where there were images of the same animal but in different trials). The top two images in Figure 4.4b are of the same animal, but experts had a majority classification of ‘lynx’ for the left image and ‘bobcat’ for the right image. Similarly, the bottom two images in Figure 4.4b are of the same animal, but experts had a majority classification of ‘unknown’ for the left image and ‘bobcat’ for the right image. In both cases, experts did not know where each image was taken, and images had the same characteristics, but the images varied slightly in the perspective of the animal.  The three expert groups, based on region, had similar levels of agreement. However, southern BC experts had slightly lower agreement than the other expert groups; nine northern BC and Yukon experts had an agreement of K = 0.63 (95% CI = 0.61-0.65), eight southern BC experts had an agreement of K = 0.59 (95% CI = 0.57-0.60), and 10 northwestern US experts had an agreement of K = 0.66 (95% = 0.64-0.67). Expert groups had different majority classifications for only 6% (n = 15) of images: 13 where one or two groups had a majority classification of ‘bobcat’ or ‘lynx’ while the other group(s) had a majority classification of ‘unknown’, and two  59 where different groups had a majority classification of ‘bobcat’ and ‘lynx’. The three expert groups had similar levels of agreement for images for which I provided locations, and also had similar consistency for retested images.  Experts did reach a clear majority classification for most images (Figure 4.2a). On average, classifications of a single expert matched the majority classification of all 27 experts for 87% of the 259 images (median = 90%, range = 64-97%). For five or more randomly selected experts, there was a >0.90 probability that their majority classification matched the final majority classification of all experts, but a probability of 0.95 required 11 or more experts (Figure 4.5). However, even majority classifications were not necessarily correct.  If the majority classification was correct for all images, then experts were incorrect in 4% of classifications, excluding classifications of ‘unknown’ (i.e., 238 out of 6,240 individual expert classifications of either ‘bobcat’ or ‘lynx’ did not match the majority classification). If the majority classification was incorrect for all images, the misclassification rate would instead be 37%. The true misclassification rate is probably somewhere between these two bounds. Although I could not conclusively determine whether expert classifications were accurate, I am nearly positive that my own classifications were accurate for 102 images based solely on their locations: 29 bobcat images were from the southern coast of BC where lynx are absent, and 73 lynx images were from north of Highway 16 in northern BC where bobcats are absent (Figure 4.1; Chapter 2). If my classifications of these images are accurate, then the majority classification of all experts was correct for all images in this subset excluding three images that had a majority classification of ‘unknown’. The misclassification rate for this subset of images would be 4%, excluding classifications of ‘unknown’ (i.e., 91 out of 2,481 individual expert classifications of either ‘bobcat’ or ‘lynx’ did not match my classification).  60  Finally, I note that I agreed with the majority classification of all 288 images in my experiment that had a majority classification of ‘bobcat’ or ‘lynx’ except for two; one majority classification by the experts was lynx, one bobcat, while I held the opposite views. One of those images is the top right image in Figure 4.4b.  Discussion  I demonstrate poor agreement among experts in distinguishing between images of two similar sympatric and congeneric species, bobcats and lynx. Experts had different levels of agreement for images with different characteristics, but in no case did experts have high enough agreement that I would consider the classification of such images by a single expert to be reliable. While I could not calculate the absolute misclassification rate because the true classifications of animals in the images were not independently confirmed, the misclassification rate was between 4% and 37%.  Further, experts classified images as ‘unknown’ in 11% of classifications, and 71% of images had ≥1 experts classify that image as ‘unknown’. Thus in many cases experts were not confident enough to classify the species in the image. These results are troubling given that the images were all of high photographic quality. I provided the option of classifying each image as ‘unknown’ rather than forcing experts to choose between ‘bobcat’ and ‘lynx’ to allow for such cases of genuine uncertainty. If I had forced experts to assign a species to each image, my calculated minimum misclassification rate of 4% would likely have been much higher. I recommend that researchers honour and trust cases of uncertainty where they cannot confidently classify the species in an image.  Experts had different levels of agreement for different kinds of images. The largest  61 difference was between summer and winter images; experts had much lower agreement for summer images. Bobcats and lynx are likely more difficult to distinguish in the summer because lynx have a much lighter summer coat and often become more brownish, and hence more similar to bobcats, whereas in the winter lynx have a thick, gray-silver coat. Experts also had lower agreement for images showing only the face or only the legs and tail than images showing the full body or only the face and legs of an animal, suggesting that it may be easiest to distinguish between the two species when both the face and legs are visible. Surprisingly, experts had slightly higher agreement for images taken at night than images taken during the day. Perhaps experts found it easier to distinguish between the two species at night because they were forced to focus on the physical features of each animal, rather than taking the colour of an animal into account, which can be misleading.  Experts also had varying levels of agreement depending on the background of images. Experts may have been cueing in on certain background features to aid in their classifications, for example associating tree species or habitat with one species over the other. Some of the experts spontaneously commented to me after the study was complete that for some images they had based their classifications on the vegetation. Experts had lower agreement for images with a forest background than images with a background of grassland or developed environments (i.e., showing human infrastructure), likely because whereas grassland and developed habitats are more characteristic of bobcats, both species use forests, thus experts could not use a forest background to help distinguish between the two species.  Further, I showed that the location of an image may also affect expert classification. Experts had greater agreement when they were provided with the location of an image. Again, spontaneous post-study comments from experts revealed that some experts used the location of  62 an image to “confirm” their selections. Surprisingly, experts had greater agreement for southern images than northern images when the location was provided. I expected the opposite because bobcats are likely absent in northern BC, thus there was essentially only one choice for northern images, whereas knowing the location of southern images should have provided little help since both species are common there (Chapter 2). Instead, many experts classified images from very northern parts of the province as ‘bobcats’, suggesting that experts were not familiar enough with the distribution of each species in BC to compare northern and southern images.  Despite these troubling results, experts did reach a clear majority classification for most images, and I believe that the majority classification was correct for most images. Thus, while classifications of an image by a single expert were unreliable, I believe that if enough experts classify an image, they usually get the classification right in the end. My findings suggest that the location of an expert did not matter, as long as many experts were asked. I found only slight differences in agreement between experts from northern BC and the Yukon, southern BC, and the northwestern US, suggesting that experts were not biased by the rarity of a species in the area where they live.   While I asked experts to classify animals based on single images, systematic camera trapping often produces multiple images of the same animal, which would likely improve classification because there would be multiple angles and visible features of the animal to observe. It would be interesting to conduct a study based on sequences of images for each animal to see how much improvement in agreement resulted from working from several images of the same animal. To maximize successful image classification, I recommend that camera-trapping studies collect multiple images for each trigger, and that citizen-science studies request all available images of an individual animal.  63  Although here I measured agreement among experts in classifying single images, my experiment was based on the best-case scenario of experts classifying images of high photographic quality. While I randomly selected images for my experiment to ensure that I did not consciously or subconsciously choose images that were easy or difficult to classify, my initial screen of using only high-quality images meant that my collection of images was likely easier to classify than images that would typically be collected in camera-trapping or citizen-science studies. I expect higher misclassification rates in such studies where it is common to encounter images of animals that are blurrier or are more distant from the camera. Misclassification rates would also likely be higher when images are classified by non-experts, such as volunteers (McShea et al. 2016) or crowd-sourcing classifications (Swanson et al. 2016). Thus, image classifications by non-experts may be suitable for species that are distinctive, but I strongly suggest caution when classifying images that contain species with similar sympatrics; I recommend that such images be flagged for later classification by multiple experts.  While image classification using automated software is not yet as accurate as manual classification (Yu et al. 2013, Gómez et al. 2016), such technology may be the future of image classification (Wearn and Glover-Kapfer 2017). However, I caution the application of automated classification for images containing species with similar sympatrics, and at the very least recommend that those images be flagged for manual classification by multiple experts until such technology is proven to reliably distinguish between images of similar species.  As camera trapping becomes increasingly common in ecological studies for many groups of species (Rowcliffe and Carbone 2008, Burton et al. 2015, Steenweg et al. 2017), I urge researchers to take a step back and evaluate how they classify their images. My findings suggest that the most basic step in the analysis of such data, species identification, may be inaccurate in  64 many cases. Reviews on the best-practices for camera trapping focus on data management and sharing (Wearn and Glover-Kapfer 2017, Scotson et al. 2017). I highlight the need to also think carefully about image classification. I recommend that studies using wildlife images consult at least five species experts when classifying images showing species with similar sympatrics. Still, I stress that the majority classification of even five experts is not necessarily correct, only that the majority classification is unlikely to change by asking more experts. When sharing metadata from camera-trapping studies, I recommend that researchers include information on the number of people that classified each image, whether those people were experts or non-experts, and the individual classifications of each person. Further, I suggest that researchers make available the raw images so that later studies have the option of reclassifying certain images.  Images have been described as being conclusive evidence for the presence of a species, even when that species is thought to be absent or extinct (McKelvey et al. 2008), but that is only true if the species in the image can be conclusively classified. I show that experts find it difficult to distinguish between images of similar species, which implies that camera-trapping or citizen-science collected images should not be taken as definitive evidence of species presence for any species that may be readily misclassified as a similar sympatric, but rather as an initial subjective screen and then followed with definitive objective survey methods such as non-invasive DNA sampling or live-trapping.      65 Table 4.1. Characteristics of the 15 image categories within the six trials.  Trial Season Background habitat Time Visible features Location provided 1) Season         Summera (n = 20) Summer Forest Day 2 of: face, legs, or tail No    Winterb (n = 20) Winter Forest Day 2 of: face, legs, or tail No 2) Background habitat         Forest (n = 20) Summer Forest Day 2 of: face, legs, or tail No    Grassland (n = 20) Summer Grassland Day 2 of: face, legs, or tail No    Developedc (n = 20) Summer Developed Day 2 of: face, legs, or tail No 3) Visible features         Full body (n = 20) Winter Forest Day Face, legs and tail No    Face only (n = 20) Winter Forest Day Face only No    Face and legs (n = 20) Winter Forest Day Face and legs only No    Legs and taild (n = 19) Winter Forest Day Legs and tail only No 4) Time         Day (n = 20) Winter Forest Day 2 of: face, legs, or tail No    Nighte (n = 20) Winter Forest Night 2 of: face, legs, or tail No 5) Location         a) Location provided            Northern BC (n = 20) Summer Forest Day 2 of: face, legs, or tail Yes       Southern BC (n = 20) Summer Forest Day 2 of: face, legs, or tail Yes    b) Location not provided            Northern BC (n = 20) Summer Forest Day 2 of: face, legs, or tail No       Southern BC (n = 20) Summer Forest Day 2 of: face, legs, or tail No 6) Consistencyf       aImages taken between April and September, and showing no snow. bImages taken between October and March, and showing snow. cImages showing human infrastructure, such as houses, barns, or patios. dOne image was mistakenly included twice in the trial; responses for the second time it appeared were removed from all analyses. eBlack and white images taken at night. fThis trial contained the same images as the first trial, but they were randomly re-ordered.  66 Table 4.2. Examples of images with poor agreement among experts in their classifications (n = 27 experts). With three classification options, the proportion of agreement had an upper bound of 1.00, indicating perfect agreement, and a lower bound of 0.31, indicating perfect disagreement. Images were cropped from original versions, thus they do not show all of the background features observed by the experts that classified them. Images provided by: A) Paul Morgan, B) Amber Piva, C) Jacqueline Brown, D) Myrna Blake, E) Bert Gregersen, F) Scott MacDonald, G) Donald Hendricks, and H) John E. Marriott.     No. of expert classifications      Image Bobcat Lynx Unknown Proportion of agreement Season Background habitat Visible features Location A 8 13 6 0.34 Summer Grassland Full body McBride B 7 15 5 0.39 Summer Developed Full body Kamloops C 6 16 5 0.41 Summer Forest Full body Tatla Lake D 8 3 16 0.43 Summer Grassland Legs and tail Fort Nelson E 5 5 17 0.44 Winter Forest Face only Vernon F 16 9 2 0.45 Summer Developed  Full body Invermere G 17 7 3 0.46 Summer Forest Full body Cache Creek H 7 19 1 0.55 Summer Grassland Full body Fort Nelson  67 Table 4.3. Agreement among all experts (n = 27) in their classifications of images within each category of images. Fleiss’ Kappa (K) measures agreement among a group of classifiers; a value of 1 indicates perfect agreement, while a value of 0 indicates agreement that would occur by chance.  Category No. of images No. of images with a unanimous classification K (95% CI) Season       Summer 20 1 0.36 (0.34-0.37)    Winter 20 6 0.77 (0.75-0.79) Background habitat       Forest 20 3 0.47 (0.45-0.49)    Grassland 20 6 0.64 (0.62-0.66)    Developed 20 10 0.66 (0.64-0.68) Visible features       Face only 20 6 0.66 (0.64-0.68)    Legs and tail 19 3 0.67 (0.65-0.69)    Full body 20 6 0.77 (0.75-0.79)    Face and legs 20 8 0.81 (0.79-0.83) Time       Day 20 2 0.58 (0.56-0.60)    Night 20 4 0.64 (0.62-0.66)     Combinations (all daytime)a       Summer, forest, legs and tail 15 0 0.34 (0.32-0.36)    Summer, developed, full body 10 6 0.40 (0.38-0.43)    Summer, forest, full body 39 6 0.47 (0.46-0.48)    Summer, forest, face and legs 24 3 0.51 (0.49-0.53)    Summer, grassland, full body 12 4 0.58 (0.56-0.60)    Winter, forest, legs and tail 26 3 0.60 (0.59-0.62)    Winter, forest, face only 20 6 0.66 (0.64-0.68)    Winter, forest, full body 36 8 0.74 (0.73-0.76)    Winter, forest, face and legs 35 14 0.80 (0.78-0.81)     Location provided       Northern BC 20 3 0.21 (0.19-0.23)    Southern BC 20 4 0.62 (0.60-0.64)    Total 40 7 0.50 (0.49-0.51) Location not provided       Northern BC 20 2 0.04 (0.02-0.06)    Southern BC 20 1 0.55 (0.53-0.57)    Total 40 3 0.44 (0.43-0.45)  aImages were pooled together from all categories excluding the 40 images for which I provided locations. Only combinations with ≥10 images are shown.   68   Figure 4.1. Images of bobcats (Lynx rufus; white circles; n = 805) and lynx (L. canadensis; black circles; n = 807) taken during 2008-2017. These images were solicited from the public across British Columbia and here I map points based on my own classifications of the images. I also show my boundary between northern and southern BC (dotted line).  69   Figure 4.2. Distribution of a) the number of experts that classified individual images as the majority classification and b) the proportion of agreement scores among all 27 experts for individual images in all categories excluding the 40 images for which I provided locations (n = 259 images). With three classification options, the proportion of agreement had an upper bound of 1.00, indicating perfect agreement, and a lower bound of 0.31, indicating perfect disagreement.  70   Figure 4.3. Agreement among all experts (n = 27) in their classifications of images within each category of images. Error bars represent 95% confidence intervals. Fleiss’ Kappa measures agreement among a group of classifiers; a value of 1 indicates perfect agreement, while a value of 0 indicates agreement that would occur by chance.   71   Figure 4.4. Examples of how the location of an image and the visible features of an animal can affect expert classification. a) Both images are of the same animal taken near Prince George, British Columbia and have the same image characteristics. The image on the left was not included in my experiment but had a 4:4 split vote between bobcat (Lynx rufus) and lynx (L. canadensis) among local biologists who were asked to classify the image. I included the image on the right in my experiment without providing its location; 26 experts classified the image as ‘bobcat’, and one expert classified the image as ‘unknown’. b) The top two images are of the same animal but show slightly varying body parts and had different majority classifications by the experts; the same occurred for the bottom two images. I show the number of experts that classified each image as ‘bobcat’, ‘lynx’, and ‘unknown’. Images provided by (from top to bottom row): James Gagnon, BC Parks, Emre Giffin.  72   Figure 4.5. Average probability that the majority classification of a randomly selected subset of experts matched the majority classification of all 27 experts, calculated from all images excluding the 40 images for which I provided locations (n = 259 images). Error bars represent one standard deviation from the mean. Probabilities are lower for even numbers of experts because of the likelihood of drawing a split vote, which is not possible for odd numbers of experts.            73 Chapter 5  Conclusions   In this thesis, I used bobcats (Lynx rufus) and Canada lynx (L. canadensis; hereafter lynx) in British Columbia, Canada, as a case study to assess the data sources available for estimating mesocarnivore distributions (Chapter 2), determine whether range shifts have occurred in the last century (Chapter 3), and determine whether similar species can be reliably classified from camera images (Chapter 4).  In Chapter 2, all data sources that I examined indicated similar distributions for each species. Bobcats were restricted to the southern half of BC, whereas lynx occurred across most of the interior of the province. Bobcat and lynx distributions broadly overlapped in southern BC, but the two species appeared to be segregated by elevation, likely due to different adaptations and resource requirements. BC provides an opportunity to study interactions between bobcats and lynx; future work should estimate co-occurrence and explore possible competition or other interactions between these two congeneric species along elevation gradients in the areas of sympatry that I have identified.  In Chapter 3, provincial trapping records indicated that ranges have remained stable over the last century for both species, and surveys from trappers supported these findings. However, climate change is accelerating (Islam et al. 2017), thus it would be interesting to see if bobcats or lynx shift their ranges in the future. These findings are very different than eastern North America where range shifts for both species have been documented (Lavoie et al. 2009, Koen et al. 2014). BC is mountainous, which may help explain the differing range dynamics here than in the far  74 flatter and lower-elevation terrain of eastern North America since climate change may be slower in mountainous regions (Loarie et al. 2009). I suspect that many wide-ranging species may show similar patterns, with some areas of the range edge that are shifting while others are stable. My findings echo previous work that showed that the ranges of many species shift in parts of their geographic ranges, but remain stable or shift in opposite directions in other parts of their ranges (Gibson-Reinemer and Rahel 2015). Future research should investigate the factors that control bobcat and lynx distributions in BC and do more work on the energetics of both species to see if climate-related factors are limiting for these species.  Lastly, in Chapter 3, I show that even experts find it difficult to classify similar-looking species from images. Overall, agreement among experts in distinguishing between bobcat and lynx images was poor (Fleiss’ Kappa = 0.64), but varied depending on different image characteristics. Agreement was higher when experts knew the location of an image versus when they did not know where an image was taken. Further, experts were inconsistent even with themselves, changing their classifications of numerous images when they were asked to reclassify the same images weeks later; on average, experts changed their classifications on seven of the 40 images. Despite these troubling results, most of the tested images did obtain a clear majority classification from the experts, although I emphasize that even majority classifications may be incorrect. These results are very clear: classification of images by a single expert is unreliable for similar-looking species. I recommend that researchers using camera-trapping images consult multiple species experts to increase confidence in their image classifications of similar sympatric species.  My work demonstrates the advantages and disadvantages of trapping records, hunting records, vehicle-kill records, trapper surveys, and images submitted by the public for estimating  75 mesocarnivore distributions. Each method has unique advantages and limitations that must be taken into account, thus the most appropriate method depends on the scale of the research question being addressed. For example, trapping records can provide broad distributions and show changes through time, but cannot be used to assess fine-scale occurrence, such as the elevations used by a species. In contrast, images submitted by the public can provide fine-scale information about elevations and habitats if the people who submit the images provide detailed descriptions of where the images were taken, but they cannot be used to assess historical distributions. Further, camera images may not be reliable for species with similar sympatrics. Trapper surveys provide both presence and absence data, but when surveys are only voluntary, participation rates are low. All other data sources do not report absence, making it difficult to interpret areas without records. I show how combining these data sources improve distribution estimates; in the case of bobcats and lynx in BC, each method contributed unique locations. More confidence that a species is absent or extremely rare can be obtained when locations lack records from all available data sources, but absence can still not be concluded.  My work also highlights the usefulness of citizen science through surveying trappers and soliciting images from the public. Active trappers across BC proved knowledgeable in assessing bobcat and lynx presence, and provided meaningful insights that complemented my analyses of harvest records. The public also showed great interest in helping to estimate bobcat and lynx distributions, and the images that they submitted comprised my second largest data set, behind only trapping records. Citizen science is becoming increasingly popular (Dickinson et al. 2012, Forrester et al. 2017, Pocock et al. 2017), and may be an important component of future mesocarnivore studies. However, I urge researchers to require images or physical evidence in order for public reports to be verifiable.  76  However, I show that even experts find it difficult to distinguish between similar species from images, thus researchers using image data in citizen science projects or camera trapping studies should also be aware of the image classification issues that I have detected. Images are thought to be strong evidence for the presence of a species, even when that species is thought to be absent or extinct, behind only a carcass or DNA evidence in terms of reliability (McKelvey et al. 2008). However, the species that the images contain must be correctly classified. My work questions the reliability of images showing species with similar sympatrics. In cases where a species with similar sympatrics is thought to be absent or extremely rare, I recommend treating camera detections as not conclusive evidence, but rather reason to survey the area to collect more concrete evidence, such as DNA.  Mesocarnivores are difficult to monitor, thus researchers and managers should examine all available data sources when estimating distributions. The limitations of each method should also be considered and addressed. While citizen science and camera trapping are becoming more common in such studies, researchers should be aware that the reliability of such records is significantly lower for similar-looking species such as bobcats and lynx, and should not be treated the same way as records of distinct-looking species.         77 References  Anderson EM, Lovallo MJ. 2003. Bobcat and lynx. Pages 758-786 in Feldhamer GA, Thompson  B. Wild Mammals of North America. Johns Hopkins University Press, Baltimore. Aubry KB, Jagger LA. 2006. The importance of obtaining verifiable occurrence data on forest  carnivores and an interactive website for archiving results from standardized surveys.  Pages 159-176 in Santos-Reis, Proulx MG, Birks JDS, O’Doherty EC. Martes in  carnivore communities. Alpha Wildlife, Sherwood Park, Alberta. Aubry KB, McKelvey KS, Copeland JP. 2007. Distribution and broadscale habitat relations of  the wolverine in the contiguous United States. Journal of Wildlife Management 71:2147- 2158. Austen GE, Bindemann M, Griffiths RA, Roberts DL. 2016. Species identification by experts  and non experts: comparing images from field guides. Scientific Reports 6:33634. BC Ministry of Forests and Range. 1998. Biogeoclimatic zones of British Columbia.  <https://www.for.gov.bc.ca/hfd/library/documents/treebook/biogeo/biogeo.htm>.  Accessed 14 Sept 2017. BC Ministry of Forests, Lands, and Natural Resource Operations. 2017. 2016-2017 Hunting and  trapping regulations synopsis. Victoria, British Columbia. BC Ministry of Transportation and Infrastructure Wildlife Accident Reporting System, © 2017  Province of British Columbia. All rights reserved. Berry JK. 1993. Beyond mapping: concepts, algorithms, and issues in GIS. Wiley, Hoboken. Brand CJ, Keith LB. 1979. Lynx demography during a snowshoe hare decline in Alberta. Journal  of Wildlife Management 43:827-849.  78 Bridger MC, Johnson CJ, Gillingham MP. 2016. Assessing cumulative impacts of forest  development on the distribution of furbearers using expert-based habitat modeling.  Ecological Applications 26:499-514. Brockie RE, Sadleir RMFS, Linklater WL. 2009. Long-term wildlife road-kill counts in New  Zealand. New Zealand Journal of Zoology 36:123-134. Burton AC, Neilson CE, Moreira D, Ladle A, Steenweg R, Fisher JT, Bayne E, Boutin S. 2015.  Wildlife camera trapping: a review and recommendations for linking surveys to  ecological processes. Journal of Applied Ecology 52:675-685. Buskirk SW, Ruggiero LF, Krebs CJ. 2000. Habitat fragmentation and interspecific competition:  implications for lynx conservation. Pages 83-100 in Ruggiero LF, Aubry KB, Buskirk  SW, Koehler GM, Krebs CJ, McKelvey KS, Squires JR. Ecology and conservation of  lynx in the United States. University Press of Colorado, Boulder. Cannon S, DeGraff J. 2009. The increasing wildfire and post-fire debris-flow threat in western  USA, and implications for consequences of climate change. Pages 177-190 in Sassa K,  Canuti P. Landslides - disaster risk reduction. Springer, Berlin. Caravaggi A, Banks PB, Burton CA, Finlay CMV, Haswell PM, Hayward MW, Rowcliffe MJ,  Wood MD. 2017. A review of camera trapping for conservation behaviour research.  Remote Sensing in Ecology and Conservation. doi:10.1002/rse2.48. Cayan DR, Kammerdiener SA, Dettinger MD, Caprio JM, Peterson DH. 2001. Changes in the  onset of spring in the western United States. Bulletin of the American Meteorological  Society 82:399-415.  Chen I, Hill JK, Ohlemüller R, Roy DB, and Thomas CD. 2011. Rapid range shift of species  associated with high levels of climate warming Science 333:1024-1026.  79 Chilelli M, Griffith B, Harrison DJ. 1996. Interstate comparisons of river otter harvest data.  Wildlife Society Bulletin 24:238-246. Choi G, Robinson DA, Kang S. 2010. Changing northern hemisphere snow seasons. Journal of  Climate 230:5305-5310. Costa H, Foody GM, Jiménez S, Silva L. 2015. Impacts of species misidentification on species  distribution modeling with presence-only data. ISPRS International Journal of Geo- Information 4:2496-2518. Day CC, Westover MD, Hall LK, Larsen RT, McMillan BR. 2016. Comparing direct and  indirect methods to estimate detection rates and site use of a cryptic semi-aquatic  carnivore. Ecological Indicators 66:230-234. DeVink JD, Berezanski D, Imrie D. 2011. Comments on Brodie and Post: harvest effort: the  missing covariate in analyses of furbearer harvest data. Population Ecology 53:261-262. Dickinson JL, Shirk J, Bonter D, Bonney R, Crain RL, Martin J, Phillips T, and Purcell K. 2012.  The current state of citizen science as a tool for ecological research and public  engagement. Frontiers in Ecology and the Environment 10:291-297. Dorendorf RR, Fix PJ, Prugh LR. 2016. Motivations of fur trappers in interior Alaska. Human  Dimensions of Wildlife 21:522-537. Elton C, Nicholson M. 1942. The ten-year cycle in numbers of the lynx in Canada. Journal  of Animal Ecology 11:215-244. Erickson DW. 1982. Estimating and using furbearer harvest information. Pages 53-60 in  Sanderson GC. Midwest furbearer management, Proceedings of the Symposium of the  43rd Midwest Fish & Wildlife Conference. Wichita, Kansas. Fleiss JL. 1971. Measuring nominal scale agreement among many raters. Psychological Bulletin  80  76:378-382. Flower A, Murdock TQ, Taylor SW, Zwiers FW. 2013. Using an ensemble of downscaled  climate projection models to assess impacts of climate change on the potential  distribution of spruce and Douglas-fir forests in British Columbia. Environmental  Science and Policy 26:63-74. Forrester TD, Baker M, Costello R, Kays R, Parsons AW, McShea WJ. 2017. Creating advocates  for mammal conservation through citizen science. Biological Conservation 208:98-105. Furnas BJ, Landers RH, Callas RL, Matthews SM. 2017. Estimating population size of fishers  (Pekania pennanti) using camera stations and auxiliary data on home range size.  Ecosphere 8:e01747. Garrote G, Ayala RP. 2015. Assessing unverified observation data used for estimating Iberian  lynx distribution. European Journal of Wildlife Research 61:801-806. Gehrt SD, Hubert GF, Ellis JA. 2002. Long-term population trends of raccoons in Illinois.  Wildlife Society Bulletin 30:457-463. Gese EM. 2001. Monitoring of terrestrial carnivore populations. Pages 372-396 in Gittleman JL,  Funk SM, MacDonald DW, Wayne RK. Carnivore conservation. Cambridge University  Press, Cambridge. Gibbon GEM, Bindermann M, Roberts DL. 2015. Factors affecting the identification of  individual mountain bongo antelope. PeerJ 3:e1303. Gibson-Reinemer DK, Rahel FJ. 2015. Inconsistent range shifts within species highlight  idiosyncratic responses to climate warming. PLoS ONE 10: e0132103. Golden HN, Christ AM, Solomon EK. 2007. Spatiotemporal analysis of wolverine Gulo gulo  harvest in Alaska. Wildlife Biology 13:68-75.  81 Gómez A, Salazar A, Vargas F. 2016. Towards automatic wild animal monitoring: identification  of animal species in camera-trap images using very deep convolutional neural networks.  Ecological Informatics 41:24-32. Gray LK, Hamann A. 2013. Tracking suitable habitat for tree populations under climate change  in western North America. Climate Change 117:289-303. Hansen K. 2007. Bobcat: master of survival. Oxford University Press, New York. Hatler DF, Beal AMM. 2003. Furbearer management guidelines: lynx. Pages 261-280 in Hatler  DF, Beal AMM. Trapper education manual, British Columbia Trappers' Association,  British Columbia. Hatler DF, Poole KG, Beal AMM. 2003. Furbearer management guidelines: bobcat. Pages 243- 260 in Hatler DF, Beal AMM. Trapper education manual, British Columbia Trappers'  Association, British Columbia. He Z, Kays R, Zhang Z, Ning G, Huang C, Han TX, Millspaugh J, Forrester T, McShea W.  2016. Visual informatics tools for supporting large-scale collaborative wildlife  monitoring with citizen scientists. IEEE Circuits and Systems Magazine 16:73:86. Hiby L, Lovell P, Pail N, Kumar NS, Gopalaswamy AM, Karanth KU. 2009. A tiger cannot  change its stripes: using a three-dimensional model to match images of living tigers and  tiger skins. Population Ecology 5:383-386. Hickling R, Roy DB, Hill JK, Fox R, Thomas CD. 2006. The distributions of a wide range of  taxonomic groups are expanding polewards. Global Change Biology 12:450-455. Hiller TL, Etter DR, Belant JL, Tyre AL. 2011. Factors affecting harvests of fishers and  american martens in northern Michigan. Journal of Wildlife Management 75:1399-1405. Hodges KE. 2000a. The ecology of snowshoe hares in northern boreal forests. Pages 117-162 in  82  Ruggiero LF, Aubry KB, Buskirk SW, Koehler GM, Krebs CJ, McKelvey KS, Squires  JR. Ecology and conservation of lynx in the United States. University Press of Colorado,  Boulder. Hodges, K.E. 2000b. Ecology of snowshoe hares in southern boreal and montane forests. Pages  163-2016 in Ruggiero LF, Aubry KB, Buskirk SW, Koehler GM, Krebs CJ, McKelvey  KS, Squires JR. Ecology and conservation of lynx in the United States. University Press  of Colorado, Boulder. Homyack JA, Vashon JH, Libby C, Lindquist EL, Loch S, McAlpine DF, Pilgrim KL. 2008.  Canada lynx-bobcat (Lynx canadensis x L. rufus) hybrids at the southern periphery of  lynx range in Maine, Minnesota and New Brunswick. The American Midland Naturalist  159:504-508. Hoving CL, Joseph RA, Krohn WB. 2003. Recent and historical distributions of Canada lynx in  Maine and the northeast. Northeastern Naturalist 10:363-382. Interagency Lynx Biology Team. 2013. Canada lynx conservation assessment and strategy, 3rd  ed. Forest Service Publication R1-13-19, USDA Forest Service, USDI Fish and Wildlife  Service, USDI Bureau of Land Management, USDI National Park Service, Missoula. Islam SU, Déry SJ, Werner AT. 2017. Future climate change impacts on snow and water  resources of the Fraser River Basin, British Columbia. Journal of Hydrometeorology  18:473-496. Jackson MM, Topp E, Gergel SE, Martin K, Pirotti F, Sitzia T. 2016. Expansion of subalpine  woody vegetation over 40 years on Vancouver Island, British Columbia, Canada.  Canadian Journal of Forest Research 46:437-443. Jiang G, et al. 2015. New hope for the survival of the Amur leopard in China. Scientific Reports  83  5:15475. Johnson SA, Walker HD, Hudson CM. 2010. Dispersal characteristics of juvenile bobcats in  south-central Indiana. Journal of Wildlife Management 74:379-385. Kang DH, Shi X, Gao H, Déry SJ. 2014. On the changing contribution of snow to the hydrology  of the Fraser River Basin, Canada. Journal of Hydrometeorology 15:1344-1365. Kapfer PM, Potts KB. 2012. Socioeconomic and ecological correlates of bobcat harvest in  Minnesota. Journal of Wildlife Management 76:237-242. Knowles N, Dettinger MD, Cayan DR. 2006. Trends in snowfall versus rainfall in the western  United States. Journal of Climate 19:4545-4559. Koehler GM, Maletzke BT, Von Kienast JA, Aubry KB, Wielgus RB, Naney RH. 2008. Habitat  fragmentation and the persistence of lynx populations in Washington State. Journal of  Wildlife Management 72:1518-1524. Koen EL, Bowman J, Lalor JL, Wilson PJ. 2014. Continental-scale assessment of the hybrid  zone between bobcat and Canada lynx. Biological Conservation 178:107-115. Koen EL, Bowman J, Murray DL, Wilson PJ. 2014. Climate change reduces genetic diversity of  Canada lynx at the trailing range edge. Ecography 37:754-762. Kosmala M, Wiggins A, Swanson A, Simmons B. 2016. Assessing data quality in citizen  science. Frontiers in Ecology and the Environment 14:551-560. Krebs CJ, Boonstra R, Boutin S, Sinclair ARE. 2001. What drives the 10-year cycle of snowshoe  hares? BioScience 51:25-35. Laliberte AS, Ripple W. 2004. Range contractions of North American carnivores and ungulates.  BioScience 54:123-138. Lavoie M, Collin P, Lemieux F, Jolicoeur H, Canac-Marquis P, Lariviere S. 2009.  84  Understanding fluctuations in bobcat harvest at the northern limit of their range. Journal  of Wildlife Management 73:870-875. Lenoir J, Svenning JC. 2013. Latitudinal and elevational range shifts under contemporary  climate change. Pages 599-611 in Levin SA. Encyclopedia of biodiversity, 2nd ed.  Academic Press, Oxford. Lenoir J, Svenning JC. 2015. Climate-related range shifts - a global multidimensional synthesis  and new directions. Ecography 38:15-28. Lewis JC. 2016. Draft periodic status review for the Lynx in Washington. Washington  Department of Fish and Wildlife, Olympia. 10 + iii pp. Lewis JC, Zielinski WJ. 1996. Historical harvest and incidental capture of fishers in California.  Northwest Science 70:291-297. Litvaitis JA, Tash JP, Stevens CL. 2006. The rise and fall of bobcat populations in New  Hampshire: relevance of historical harvests to understanding current patterns of  abundance and distribution. Biological Conservation 128:517-528. Litvaitis JA, Clark AG, Hunt JH. 1986. Prey selection and fat deposits of bobcats (Felis rufus)  during autumn and winter in Maine. Journal of Mammalogy 67:389-392. Liu X, Wu P, Shao X, Songer M, Cai Q, He X, Zhu Y. 2017. Diversity and activity patterns of  sympatric animals among four types of forest habitat in Guanyinshan Nature Reserve in  the Qinling Mountains, China. Environmental Science and Pollution Research 24:16465- 16477. Loarie SR, Duffy PB, Hamilton H, Asner GP, Field CB, Ackerly DD. 2009. The velocity of  climate change. Nature 462:1052-1055. Long RA, Mackay P, Zielinski WJ, Ray JC. 2008. Noninvasive survey methods for carnivores.  85  Island Press, Washington. Marston MA. 1942. Winter relations of bobcats to white-tailed deer in Maine. Journal of  Wildlife Management 6:328-337. Masek JG, Cohen WB, Leckie D, Wulder MA, Vargas R, de Jong B, Healey S, Law B, Birdsey  R, Houghton RA, Mildrexler D, Goward S, Smith BW. 2011. Recent rates of forest  harvest and conversion in North America. Journal of Geophysical Research 116:G00K03. McDonald RA, Harris S. 1999. The use of trapping records to monitor populations of stoats  Mustela erminea and weasels M. nivalis: the importance of trapping effort. Journal of  Applied Ecology 36:679-688. McKelvey KS, Aubry KB, Ortega YK. 2000. History and distribution of lynx in the contiguous  United States. Pages 207-264 in Ruggiero LF, Aubry KB, Buskirk SW, Koehler GM,  Krebs CJ, McKelvey KS, Squires JR. Ecology and conservation of lynx in the United  States. University Press of Colorado, Boulder. McKelvey KS, Aubry KB, Schwartz MK. 2008. Using anecdotal occurrence data for rare or  elusive species: the illusion of reality and a call for evidentiary standards. BioScience  58:549-555. McShea MJ, Forrester T, Costello R, He Z, Kays R. 2016. Volunteer-run cameras as distributed  sensors for macrosystem mammal research. Landscape Ecology 31:55-66. Meehl GA, et al. 2007. Global climate projections. Pages 747-845 in Quin D, Manning M, Chen  Z, Marquis M, Averyt KB, Tignor M, Miller HL. Climate change 2007: the physical  science basis. Cambridge University Press, Cambridge. Miller DA, Nichols JD, McClintock BT, Campbell Grant EH, Bailey LL, Weir, LA. 2011.  Improving occupancy estimation when two types of observational error occur: non- 86  detection and species misidentification. Ecology 92:1422-1428. Molinari-Jobin A, et al. 2012. Monitoring the presence of species misidentification: the case of  the Eurasian lynx in the Alps. Animal Conservation 15:266-273. Mote PW, Hamlet AF, Clark MP, Lettenmaier DP. 2005. Declining mountain snowpack in  western North America. Bulletin of the American Meteorological Society 86:39-44. Mowat G, Poole KG, O’Donoghue M. 2000. Ecology of lynx in northern Canada and Alaska.  Pages 265-306 in Ruggiero LF, Aubry KB, Buskirk SW, Koehler GM, Krebs CJ,  McKelvey KS, Squires JR. Ecology and conservation of lynx in the United States.  University Press of Colorado, Boulder. Murdock TQ, Taylor SW, Flower A, Mehlenbacher A, Montenegro A, Zwiers FW, Alfaro R,  and Spittlehouse DL. 2013. Pest outbreak distribution and forest management impacts in  a changing climate in British Columbia. Environmental Science and Policy 26:75-89. Murray DL, Steury TD, Roth JD. 2008. Assessment of Canada lynx research and conservation  needs in the southern range: another kick at the cat. Journal of Wildlife Management  72:1463-1472. Murray DL, Boutin S. 1991. The influence of snow on lynx and coyote movements: does  morphology affect behaviour? Oecologia 88:463-469. Najafi MR, Zwiers F, Gillett N. 2017. Attribution of the observed spring snowpack decline in  British Columbia to anthropogenic climate change. Journal of Climate 30:4113-4130. Newbury RK. 2013. Behavioural ecology of the bobcat in a region with deep winter snows.  Thesis, University of British Columbia Okanagan, Kelowna, British Columbia, Canada. O’Connell AF, Nichols JD, Karanth KU. 2011. Camera Traps in Animal Ecology. Springer, New  York.  87 Palma L, Pedro B, Rodrigues M. 1999. The use of sighting data to analyse Iberian lynx habitat  and distribution. Journal of Applied Ecology 36:812-824. Parker GR, Maxwell JW, Morton LD, Smith GEJ. 1983. The ecology of the lynx (Lynx  canadensis) on Cape Breton Island. Canadian Journal of Zoology 61:770-786. Parmesan C, Yohe G. 2003. A globally coherent fingerprint of climate impacts across natural  systems. Nature 421:37. Peers MJL, Thornton DH, Murray DL. 2013. Evidence for large-scale effects of competition:  niche displacement in Canada lynx and bobcat. Proceedings of the Royal Society B  280:20132495. Pocock MJO, Tweddle JC, Savage J, Robinson LD, Roy HE. 2017. The diversity and evolution  of ecological and environmental citizen science. PLoS ONE 12:e0172579. Poole KG, Mowat G. 2001. Alberta furbearer harvest data analysis. Alberta Sustainable  Resource Development, Fish and Wildlife Division, Alberta Species at Risk Report No.  31. Edmonton. 51 pp. Poole KG. 2003. A review of the Canada lynx, Lynx canadensis, in Canada. Canadian Field- Naturalist 117:360-376. Pozzanghera CB, Sivy KJ, Lindberg MS, Prugh LR. 2016. Variable effects of snow conditions  across boreal mesocarnivore species. Canadian Journal of Zoology 94:697-705. Reed GC, Litvaitis JA, Ellingwood M, Tate P, Broman DJA, Sirén APK, Carroll RP. 2017.  Describing habitat suitability of bobcats (Lynx rufus) using several sources of information  obtained at multiple spatial scales. Mammalian Biology 82:17-26. Rich LN, et al. 2017. Assessing global patterns in mammalian carnivore occupancy and richness  by integrating local camera trap surveys. Global Ecology and Biogeography 26:918-919.  88 Rich LN, Kelly MJ, Sollmann R, Noss AJ, Maffei L, Arispe RL, Paviolo A, De Angelo CD, Di  Blanco YE, Di Bitetti MS. 2014. Comparing capture-recapture, mark-resight, and spatial  mark-resight models for estimating puma densities via camera traps. Journal of  Mammalogy 95:382-391. Roberts DL, Elphick CS, Reed JM. 2009. Identifying anomalous reports of putatively extinct  species and why it matters. Conservation Biology 24:189-196. Roberts NM, Crimmins SM. 2010. Bobcat population status and management in North America:  evidence of large-scale population increase. Journal of Fish and Wildlife Management  1:169-174. Robichaud CB, Boyce MS. 2010. Spatial and temporal patterns of wolf harvest on registered  traplines in Alberta, Canada. Journal of Wildlife Management 74:635-643. Rowcliffe JW, Carbone C. 2008.Surveys using camera traps: are we looking to a brighter future?  Animal Conservation 11:185-186. Roy HE, Baker E, Saunders A, Pocock MJO. 2016. Focal plant observations as a standardised  method for pollinator monitoring: opportunities and limitations for mass participation  citizen science. PLoS ONE 11:e050794. Royle JA, Link WA. 2006. Generalized site occupancy models allowing for false positive and  false negative errors. Ecology 87:835-841. Schmidt JI, Kellie KA, Chapin FS. 2015. Detecting, estimating, and correcting for biases in  harvest data. Journal of Wildlife Management 79:1152-1162. Schwartz MK, Pilgrim KL, McKelvey KS, Lindquist EL, Claar JJ, Loch S, Ruggiero LF. 2004.  Hybridization between Canada lynx and bobcats: genetic results and management  implications. Conservation Genetics 5:349-355.  89 Scotson L, et al. 2017. Best practices and software for the management and sharing of camera  trap data for small and large scale studies. Remote Sensing in Ecology and Conservation  3:158-172. Sielecki LE. 2004. WARS 1983-2002: wildlife accident reporting and mitigation in British  Columbia: a special annual report. ISSN 0-7726-5222-8. Sielecki LE. 2005. Comprehensive monitoring of wildlife mortality on British Columbia  highways using the WARS system (1978 to 2005). Proceedings of the symposium on  wild animals and traffic accidents: monitoring, analysis, prevention measures, and  measure evaluation. IXth International Mammalogical Congress (IMC 9). Sapporo,  Japan. Statistics Canada. 2012. Table 003-0013 - Number and value of pelts produced, annual,  CANSIM (database). Steenweg R, et al. 2017. Scaling-up camera traps: monitoring the planet’s biodiversity with  networks of remote sensors. Frontiers in Ecology and the Environment 15:26-34. Stewart FEC, Heim NA, Clevenger AP, Paczkowski J, Vople JP, Fisher, JT. 2016. Wolverine  behavior varies spatially with anthropogenic footprint: implications for conservation and  inferences about declines. Ecology and Evolution 6:1493-1503. Swanson A, Kosmala M, Lintott C, Packer C. 2016. A generalized approach to producing,  quantifying, and validating citizen science data from wildlife images. Conservation  Biology 30:520-531. Tumlison R, McDaniel VR. 1986. Harvest trends of the bobcat (Lynx rufus) in Arkansas.  Proceedings of the Arkansas Academy of Science 40:78-81. US Fish and Wildlife Service. 2000. Endangered and threatened wild-life and plants:  90  determination of threatened status for the contiguous U.S. distinct population segment of  the Canada lynx and related rule. Final rule. Federal Register. 65:16052-16086. US Fish and Wildlife Service. 2005. Recovery outline: contiguous United States distinct  population segment of Canada lynx. US Fish and Wildlife Service, Helena. Vanbianchi CM, Gains WL, Murphy MA, Pither J, Hodges KE. 2017. Habitat selection by  Canada lynx: making due in heavily fragmented landscapes. Biodiversity and  Conservation 26:3343-3361. Wearn OR, Glover-Kapfer P. 2017. Camera-trapping for conservation: a guide to best-practices.  WWF Conservation Technology Series 1(1). WWF-UK, Woking. Webb SM, Davidson DJ, Boyce MS. 2008. Trapper attitudes and industrial development on  registered traplines in west-central Alberta. Human Dimensions of Wildlife 13:115-126. Weingarth K, Heibl C, Knauer F, Zimmermann F, Bufka L, Heurich M. 2012. First estimation of  Eurasian lynx (Lynx lynx) abundance and density using digital cameras and capture- recapture techniques in a German national park. Animal Biodiversity and Conservation  35:197-207. Woolf A, Hubert Jr GF. 1998. Status and management of bobcats in the United States over three  decades: 1970s-1990s. Wildlife Society Bulletin 26:287-293. Woolf A, Nielsen CK, Weber T, Gibbs-Kieninger TJ. 2002. Statewide modeling of bobcat, Lynx  rufus, habitat in Illinois, USA. Biological Conservation 104:191-198. Yu X, Wang J, Kays R, Jansen PA, Wang T, Huang T. 2013. Automated identification of animal  species in camera trap images. EURASIP Journal on Image and Video Processing 52.    91 Appendices  Appendix A: Bobcat and lynx harvest regimes during 1983-2015 in British Columbia.  Table A.1. Bobcat (Lynx rufus) and lynx (L. canadensis) hunting and trapping seasons, bag limits, and compulsory inspection and reporting requirements during 1983-2015 in British Columbia.   Open regions Season datesa Bag limitb CIc CRd Bobcat         Hunting 2e, 3e, 4, 5e, 8 Nov 15-Feb 15 Five None All    Trapping 2, 3, 4, 5, 8 Nov 15-Feb 15 Nonef None 2, 4, 8g Lynx         Hunting 3e, 4, 5e, 6e, 7A, 7B, 8 Nov 15-Feb 15 One 4h Alli    Trapping 3, 4, 5, 6, 7A, 7B, 8 Nov 15-Feb 15 Nonej None 4, 8k  aSeason dates varied slightly around Nov 15-Feb 15 between Regions and years. bProvincial bag limit; Regional limits also applied and have changed through time. cCompulsory inspection. dCompulsory reporting; not required prior to 1986. eSome Management Units were closed in these Regions. fFrom 2000-2011 there was a bag limit of two in Region 4. gRequired in only Region 4 during 1986-1998. hNot required prior to 2012. iExcept for Region 4 during 2012-2015. jThere was a bag limit of one in three years in Regions 4 and 8 in 1993, and in Region 4 in 1994. kRequired in only Region 4 during 1986-1993.            92 Appendix B: Wildlife Management Units and registered trapline system in British Columbia.    Figure B.1. Wildlife Management Units in British Columbia, Canada. BC contains nine Regions (thick borders) divided into 225 Wildlife Management Units (thin borders) ranging from 465 km² to 18,982 km² (   = 4,216 km²).  93   Figure B.2. Registered trapline system in British Columbia. BC currently contains 2,451 registered traplines ranging from 0.6 km² to 23,310 km² (   = 366 km²). Traplines cover most of BC except for urban centers and some provincial and national parks (gray areas).               94 Appendix C: Trapper survey. Below is the survey that I distributed to trappers across British Columbia in fall 2015 as part of my work in Chapters 2 and 3. I hand-wrote the registered trapline number above the blank line in the first paragraph on each individual survey prior to mailing them out.  BC Bobcat and Lynx Trapper Survey  Principal Investigator       Co-Investigator Dr. Karen Hodges       TJ Gooliaff  Associate Professor, Biology        MSc Student, Biologist in Training UBCO         UBCO karen.hodges@ubc.ca      tj.gooliaff@ubc.ca          This survey has been developed and funded by the University of British Columbia Okanagan as part of my master’s research. The purpose of this survey is to learn more about where bobcats and lynx occurred in BC between 2000 and 2016. I have identified your trapline as one with high activity during this time period, and thus I have selected you to receive this survey as the registered trapper of                                            . If you are not the registered trapper of this trapline, please pass this survey on to the appropriate person if possible.  Your participation in this study is entirely voluntary and you may refuse to participate or withdraw from this study at any time, although I would greatly appreciate it if you take the time to complete this survey. Your local expertise would be a great contribution to our understanding of bobcats and lynx in BC.  If you participate in this study, there will be no greater risks than what you would experience in your daily life. The information you provide will be used in my thesis, which will be publically available on the internet via cIRcle, and subsequent publications. Our findings will also be presented at the 2017 BCTA Convention and in the BC Trapper magazine. Completed surveys will remain confidential and will be securely stored at UBCO for five years after publication. After this time, surveys will be shredded. Please be assured that the results of this study will not influence hunting/trapping bag limits or season dates.  Please answer questions 1 through 9 and return this survey to the mailing address at the bottom of page 4 by January 1, 2017. Alternatively, you can email me a scanned copy, or request a PDF version of this survey instead. I anticipate this survey will take less than 30 minutes to complete. If you have any questions about this study, please don’t hesitate to send me an email or give me a call. You can also contact my supervisor Karen Hodges.  If you have any concerns or complaints about your rights as a research participant and/or your experiences while participating in this study, contact the Research Participant Complaint Line in the UBC Office of Research Services toll free at 877-822-8598 or the UBC Okanagan Research Services Office at 250-807-8832.  95 1) What year did you first start trapping this trapline?       2) Please complete the following table to the best of your knowledge and base your answers on your trapline. Don’t worry if you can’t provide information for all years - even one or two years will be useful. For seasons during which you did not trap your trapline, or spent little time on your trapline, please select N/A. Season refers to the trapping season (November through February). Use a combination of harvest totals, sightings, and sign (tracks, scat, etc) to indicate the abundance of each species on your trapline as follows:   Absent - the species did not occur on your trapline  Few - the species occurred on your trapline, but was rare  Plentiful - the species was common on your trapline  N/A - you are unable to assess relative abundance   Bobcat Lynx Season Absent Few Plentiful N/A Absent Few Plentiful N/A 2000-2001 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2001-2002 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2002-2003 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2003-2004 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2004-2005 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2005-2006 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2006-2007 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2007-2008 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2008-2009 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2009-2010 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2010-2011 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2011-2012 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2012-2013 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2013-2014 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2014-2015 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ 2015-2016 ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝ ⃝            96 3) Since you first started trapping your trapline, do you feel that the bobcat population on your trapline has increased, decreased, or stayed the same? Select N/A if bobcats are not found on your trapline.       ⃝ Increased  ⃝ Decreased  ⃝ Stayed the same  ⃝ N/A  4) Do you feel that the lynx population on your trapline cycles? Select N/A if lynx are not found on your trapline, or if you have not trapped long enough to notice any cycles.   ⃝ Yes  ⃝ No  ⃝ N/A  If yes, what trapping season marks the last high point in your lynx cycle?     5) Since you first started trapping your trapline, do you feel that the lynx population on your trapline has increased, decreased, or stayed the same? If lynx cycle on your trapline, compare low points to low points and high points to high points. For example, if you first started trapping this trapline during a high point in the cycle, compare that lynx population with the population during the last high point. Select N/A if lynx are not found on your trapline.   ⃝ Increased  ⃝ Decreased  ⃝ Stayed the same  ⃝ N/A  6) Since you first started trapping your trapline, do you feel that bobcats and lynx on your trapline now occur at higher elevations, lower elevations, or at the same elevations? Select N/A if the species is not found on your trapline.            Bobcat Higher   ⃝                       Lower   ⃝                        Same   ⃝                      N/A   ⃝      Lynx Higher   ⃝                       Lower   ⃝                        Same   ⃝                      N/A   ⃝  7) Which statement best describes the elevations at which bobcats and lynx are currently found on your trapline? If only one species is found on your trapline, select N/A.   ⃝   Bobcats are generally at lower elevations than lynx  ⃝   Bobcats are generally at higher elevations than lynx  ⃝   Bobcats and lynx are generally at the same elevations  ⃝   N/A  97 8) Please provide any additional information about bobcats and lynx on your trapline, especially about abundance, distribution, and interactions between the two species.    9) May I contact you if I have any follow-up questions about this survey? If so, please fill in your contact information below. If not, leave this question blank. Your name and contact information will remain confidential if provided.  Name:  Phone number:  Email address:    Please return this completed survey to the address below by January 1st, 2017. If the survey is completed, it will be assumed that consent has been given.  Thank You!   TJ Gooliaff  Biology Department  SCI 149, 1177 Research Road  Kelowna, BC V1V 1V7    98 Appendix D: Wanted poster for bobcat and lynx images. Below is the poster that was advertised in hunting and outdoor stores across British Columbia to help solicit bobcat and lynx images from the public as part of my work in Chapter 2.    99 Appendix E: Active traplines during 2000-2013 in British Columbia.    Figure E.1. Active traplines during 2000-2013 in British Columbia. Shown are traplines that harvested any species in ≥5 years during 2000-2013 and harvested ≥1 bobcat (Lynx rufus), lynx (L. Canadensis), red fox (Vulpes vulpes), coyote (Canis latrans), wolf (C. lupus), or wolverine (Gulo gulo) during that time (n = 796). Registered trappers of black traplines completed the survey (n = 157), whereas trappers of gray traplines were not sent the survey because their contact information was not available (n = 121), or were sent the survey but did not complete it prior to the end of my study (n = 518).           100 Appendix F: Elevations of bobcat and lynx images taken during 2008-2017 in British Columbia separated by season.    Figure F.1. Elevations of bobcat (Lynx rufus; black bars) and lynx (L. canadensis; gray bars) images taken during 2008-2017 in British Columbia estimated from all images taken during a) April 1 - September 30, and b) October 1 - March 31. Images were collected using citizen science, and elevations were estimated from location descriptions provided by the people who submitted the images. Location details varied in quality and were often vague, thus these values are coarse estimates. Bobcat and lynx sketches were courtesy of N. Reynolds at Into The Wild Artistry.  101 Appendix G: Distributions of bobcats and lynx from trapping records summarized by Management Units in British Columbia each year during 1983-2013.    102   Figure G.1. Distribution of bobcats (Lynx rufus) from trapping records summarized by Management Units in British Columbia each year during 1983-2013. Black Management Units harvested ≥1 bobcats during each year. Gray Management Units contained active trappers, but had no harvest of bobcats. White areas contained no active trappers.  103   104   Figure G.2. Distribution of lynx (Lynx canadensis) from trapping records summarized by Management Units in British Columbia each year during 1983-2013. Black Management Units harvested ≥1 lynx during each year. Gray Management Units contained active trappers, but had no harvest of lynx. White areas contained no active trappers.     105 Appendix H: Northern-most confirmed bobcat in British Columbia.    Figure H.1. Northern-most confirmed bobcat (Lynx rufus) in British Columbia. This female bobcat was trapped near Houston, BC, in December 2015, and was given to BC’s Ministry of Forests, Lands, and Natural Resource Operations because there is no trapping season for bobcats in that area. Genetic analysis confirmed that the animal was a pure bobcat (R. Weir, BC Ministry of Environment, unpublished data). Images taken by TJ Gooliaff in April 2016.               106 Appendix I: Online image classification experiment. I used an online survey tool to ask 27 experts to classify 300 images as ‘bobcat’, ‘lynx’, or ‘unknown’ as part of my work in Chapter 4.       Figure I.1. Typical screen shot of one of the image pages in one of the six online image classification trials. Image provided by Donald Hendricks.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0362863/manifest

Comment

Related Items