Patterns of River Otter {Lontra canadensis) Diet and Habitat Selection at Latrine Sites in Central British Columbia Shannon Michael Crowley B.Sc, University of Alaska Southeast, 1999 Thesis Submitted in Partial Fulfillment of The Requirements for the Degree of Master of Science In Natural Resources and Environmental Studies (Biology) The University Of Northern British Columbia August 2009 ©Shannon Michael Crowley, 2009 1*1 Library and Archives Canada Bibliotheque et Archives Canada Published Heritage Branch Direction du Patrimoine de I'edition 395 Wellington Street OttawaONK1A0N4 Canada 395, rue Wellington Ottawa ON K1A 0N4 Canada Your file Votre reference ISBN: 978-0-494-60815-9 Our file Notre reference ISBN: 978-0-494-60815-9 NOTICE: AVIS: The author has granted a nonexclusive license allowing Library and Archives Canada to reproduce, publish, archive, preserve, conserve, communicate to the public by telecommunication or on the Internet, loan, distribute and sell theses worldwide, for commercial or noncommercial purposes, in microform, paper, electronic and/or any other formats. L'auteur a accorde une licence non exclusive permettant a la Bibliotheque et Archives Canada de reproduire, publier, archiver, sauvegarder, conserver, transmettre au public par telecommunication ou par Nnternet, preter, distribuer et vendre des theses partout dans le monde, a des fins commerciales ou autres, sur support microforme, papier, electronique et/ou autres formats. The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. L'auteur conserve la propriete du droit d'auteur et des droits moraux qui protege cette these. Ni la these ni des extraits substantiels de celle-ci ne doivent etre imprimes ou autrement reproduits sans son autorisation. In compliance with the Canadian Privacy Act some supporting forms may have been removed from this thesis. Conformement a la loi canadienne sur la protection de la vie privee, quelques formulaires secondaires ont ete enleves de cette these. While these forms may be included in the document page count, their removal does not represent any loss of content from the thesis. Bien que ces formulaires aient inclus dans la pagination, il n'y aura aucun contenu manquant. •+l Canada Patterns of River Otter {Lontra canadensis) Diet and Habitat Selection at Latrine Sites in Central British Columbia Shannon Michael Crowley ABSTRACT I used an Information Theoretic Model Comparison approach to investigate the relationships among river otter {Lontra canadensis) diet and temporal/spatial parameters and habitat characteristics and the presence, consistency, and intensity of otter activity. Data were collected every two weeks at latrine sites used by otters in central British Columbia from 2007-2008. In general, a stable-isotope analysis agreed with the results of a scat content analysis showing a dominance of fish in the diet of otter and a small contribution from other prey sources. I generated predictive models for the presence of salmonids, minnows, and insects in otter scat and for number of otter scats at latrine sites. For latrine selection, I found that habitat characteristics at the fine scale were better at predicting the presence of latrine sites. In general, otter activity was influenced by parameters that described vegetation cover at the fine scale and by characteristics of aquatic habitat beneficial to fish at the coarse scale. TABLE OF CONTENTS Chapter 1: General Introduction Background Study Area 1 2 7 Chapter 2: Spatio-temporal variation in river otter diet and latrine site activity on Tezzeron and Pinchi Lakes, British Columbia Abstract Introduction Methods Data Collection Stable-isotope Analysis Diet and Scat Model Development Diet and Scat Model Selection Predictive Ability of Diet and Scat Models Results Scat content analysis Stable Isotopes Diet Models Scat Deposition Models Discussion 11 12 13 18 18 22 23 28 28 29 29 33 37 42 46 Chapter 3: Spatial and behavioural scales of habitat selection and activity by river otters at latrine sites Abstract Introduction Methods Data Collection Fine-scale Selection and Activity Coarse-scale Selection and Activity Habitat Models Selection of Latrine sites Latrine Consistency Latrine Intensity Habitat Model Development Habitat Model Selection Predictive Ability of Habitat Models Results Selection of Latrine Sites Latrine Consistency Latrine Intensity Discussion 56 57 58 62 62 63 64 65 66 67 67 67 72 73 74 74 75 75 84 IV Chapter 4: General Summary References 93 98 v LIST OF FIGURES Figure 1. Map of river otter study area and John Prince Research Forest including locations of latrine sites, 2007-2008 10 Figure 2. Percent frequency of occurrence of prey groups in river otter scats on and near Tezzeron and Pinchi Lakes, central British Columbia from late May to October, 2007-2008 30 Figure 3. Percent frequency of occurrence of fish prey groups in river otter scats on and near Tezzeron and Pinchi Lakes, central British Columbia from late May to October, 2007-2008 (E = early, L = Late) 31 Figure 4. Percent frequency of occurrence of non-fish prey items in river otter scats on and near Tezzeron and Pinchi Lakes, central British Columbia from late May to October, 2007-2008 (E = early, L = Late) 32 Figure 5. Mean (+ SE) stable isotopic signatures for potential prey items of river otters on Tezzeron and Pinchi Lakes, central British Columbia 35 Figure 6. Comparison of the best ZINB model predictions versus observed probabilities of counts of river otter scats at latrine sites on Tezzeron and Pinchi Lakes, central British Columbia from May to October (2007-2008). Spatio-temporal variables (geographic zone, two-week time period) were used to predict the number of scats at latrine sites 45 Figure 7. Predicted versus observed probability of scat counts for river otter on Tezzeron and Pinchi Lakes, British Columbia, from May to October (2007-2008). Predictions were generated with the best fine-scale ZINB model and an independent data set 82 Figure 8. Predicted versus observed probability of scat counts for river otter on Tezzeron and Pinchi Lakes, British Columbia, from May to October (2007-20008). Predictions were generated with the best coarse-scale ZINB model and an independent data set 83 VI LIST OF TABLES Table 1. Fish family groups identified in river otter scats on Tezzeron and Pinchi Lakes, central British Columbia during the ice-free season (2007-2008) 21 Table 2. Parameters used in the development of binary predictive models for presence of river otter prey items and for ZINB count predictive models of scats on Tezzeron and Pinchi Lakes in central British Columbia, from 2007-2008 25 Table 3. A priori candidate models for the binary (prey presence) and count analysis (number of scats) of river otter diet and scat deposition on Tezzeron and Pinchi Lakes in central British Columbia, based on data collected from 2007-2008 27 Table 4. Estimated range (%) of prey sources in river otter diet on Tezzeron and Pinchi Lakes, central British Colubmbia, from 2007-2008. Estimates generated by analysis of stable isotopes using a dietary mixing model (Isosource) including a sensitivity analysis of 15% changes in fractionation values for carbon (AC) and nitrogen (AN) 36 Table 5. Summary of AICC model selection statistics for candidate binary models used to predict prey occurrence in river otter scats on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008. Results provided only for prey group models with good predictive performance 38 Table 6. Estimated coefficients for AICC selected models (binary) describing the presence of prey items in river otter diet on Tezzeron and Pinchi Lakes, central British Columbia (2007-2008) 40 Table 7. Summary of AICC model selection statistics for candidate ZINB models to predict counts of river otter scats using spatio-temporal variables. Data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008 41 Table 8. Estimated coefficients for the single model (ZINB count) with the lowest AICC score describing the effects of spatio-temporal variables on intensity of activity at latrine sites by river otter. Data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008 43 Table 9. Parameters used in the development of binary and ZINB count models for the selection of latrine sites and activity by river otter, based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia, from 2007-2008 69 vn Table 10. Parameters used in the development of binary and ZINB count models for the selection of latrine sites and activity by river otter, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia, from 2007-2008 70 Table 11. A priori candidate models for the selection of latrine sites and activity (presence, consistency, and intensity) by river otter, based on fine- and coarse-scale data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008 71 Table 12. Summary of AICC model selection statistics for candidate models (binary and ZINB count) predicting latrine selection and activity (occurrence, consistency, and intensity), based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008 76 Table 13. Estimated coefficients for AICC selected models (binary) predicting the selection of latrine sites and consistency of activity by river otters, based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008 77 Table 14. Summary of AICC model selection statistics for candidate models predicting latrine activity (consistency and intensity) by river otter, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008 78 Table 15. Estimated coefficients for the AICC selected model (binary) predicting the consistency of activity by river otters at latrine sites, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008 78 Table 16. Estimated coefficients for the AICC selected model (ZINB count) predicting the intensity of activity of by river otters latrine sites, based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008 79 Table 17. Estimated coefficients for the AICcselected model (ZINB count) predicting the intensity of activity of latrine sites by river otters, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008 81 vm LIST OF APPENDICES Appendix 1. Stable isotopic signature means, standard error, and 95% confidence intervals for potential prey items of river otter (Lontra canadensis) on Tezzeron and Pinchi Lakes, central British Columbia (2007-2008) 107 Appendix 2. Detailed description of fine-scale habitat measurements at river otter latrine sites on Tezzeron and Pinchi Lakes, central British Columbia (2007-2008) 108 IX ACKNOWLEDGEMENTS The author wishes to thank the Habitat Conservation Trust Fund, Tech Cominco, the John Prince Research Forest, and the University of Northern British Columbia for providing funding to support this project. Dr. Walter Wigmore, Dr. Elena Garde, and Dr. Malcolm McAdie provided excellent veterinary care for captured otters. Eric Stier and other pilots at Guardian Aerospace provided their skill and patience during telemetry flights. Dr. Helen Schwantje provided her support and advice in handling otters. I would like to thank Bob Fredericks, President of the Fort St. James Chapter of the BCTA, for providing the hair samples necessary for stable-isotope analysis. John Orlowsky loaned me a dissecting microscope for scat identification for much longer than either of us had expected or hoped. I would like to thank my supervisor, Dr. Chris Johnson, for his support and assistance throughout the entire project. I would also like to thank my committee members, Dr. Michael Gillingham and Douglas Heard, for their interest and guidance in this project. I thank Sue Grainger, Beverly John, and Amelia Stark for their support and shared space. A special thanks to Sarah Champagne and Nathan Tom for their hard work, positive attitude, and sense of humour. I thank Sebastian Anatole for his great companionship at the field site and for both his solicited and unsolicited advice. I would especially like to thank Dexter Hodder for his help, generosity, and friendship throughout the project and beyond. I especially thank the two most important people in my life, Valerie and Willa, for their support, reality checks, and love at the end of each day. x CHAPTER 1 General Introduction General Introduction Background River otters (Lontra canadensis) are distributed in freshwater and marine systems throughout North America. Otters are a top-level predator in aquatic ecosystems and are considered a "sentinel" species, or one that is sensitive to environmental disturbance and useful for measuring or indexing ecosystem health (Bowyer et al. 2003). Declines of otter populations in the US and Europe {Lutra lutra) support this designation and suggest that otters are sensitive to a range of environmental disturbances that include over harvest, habitat destruction, and pollution (Foster-Turley et al. 1990; Raesly 2001; Melquist et al. 2003). Previous studies suggest that the diet of river otter consists predominately of fish species (Melquist and Hornocker 1983; Larsen 1984; Reid et al. 1994b). Insects, clams, snails, snakes, turtles, waterfowl, shore birds, beaver, and muskrat have been documented as secondary components of otter diets (Toweill 1974; Melquist and Hornocker 1983; Larsen 1984; Stenson et al. 1984; Reid et al. 1994b). Typically, the diet of otters varies seasonally (Melquist and Hornocker 1983; Reid et al. 1994b). Scat or fecal analysis is the predominant technique for identifying the diet of river otter populations. Collection of scats is facilitated by an otter's tendency to use latrine sites, or identifiable and consistent areas on land where otter deposit their feces and concentrate their activity. Bowyer et al. (2003) suggested that river otters may be a keystone species because they transport nutrients into terrestrial systems when defecating at latrine sites. The transport of these nutrients can shape community composition in the near shore environment (Ben-David et al. 1998). 2 The distribution of river otter populations has been described for both marine and freshwater systems. A small number of studies employed radio-telemetry techniques to investigate both movements and habitat use (Melquist and Hornocker 1983; Reid et al. 1994a; Gorman et al. 2006; Helon 2006). Otters are a wide-ranging species with adult male home ranges typically larger than adult females (Gorman et al. 2006). Reid et al. (1994a) documented minimum convex polygon home ranges from 15.8 km for an adult female with young to 271.9 km2 for an adult male. Otters have also been documented traveling and feeding in both family groups and groups of young males (Reid et al. 1994b; Blundell et al. 2002). In a freshwater system, Melquist and Hornocker (1983) suggested that food has the greatest influence on the frequency and extent of movements by otters. The authors documented seasonal movements by otters to abundant food resources such as kokanee (Oncorhynchus nerka) spawning grounds. In Yellowstone Lake, otters also appeared to make movements in direct relation to the timing of spawning cutthroat trout {Oncorhynchus clarki) (Crait and Ben-David 2006). Telemetry locations from otter in Alberta suggest strong seasonal habitat selection for shoreline morphology and substrate (Reid et al. 1994a). At the scale of the annual home range, the strongest selection for habitats occurred in winter, followed by the breeding season, and then the ice-free season. The most strongly selected habitats contained stream sections with beaver ponding or lakeshores with silt or organic substrate. Tributaries, points of land, coniferous trees, rock formations, and fallen logs often characterize latrine sites (Dubuc et al. 1990; Newman and Griffin 1994; Swimley et al. 1998). Beaver activity has also been reported as an important factor influencing otter distribution (Melquist and 3 Hornocker 1983; Dubuc et al. 1990; Swimley et al. 1998). Habitat selection by mammals is strongly correlated with food resources (Costello and Sage 1994; Johnson et al. 2001), however, there is very little information documenting variation in otter diet and habitat selection relative to the seasonal availability of prey species. River otter populations in North America declined dramatically throughout the 1800s and early 1900s. By the late 1970s, otter populations occurred across less than 75% of their historical distribution (Melquist et al. 2003). Pollution was one of the leading factors that contributed to the historical decline of river otter. The otter's trophic level in an aquatic system makes it especially vulnerable to heavy-metal contamination (Kimber and Kollias 2000). Another factor that contributed to the historical decline of otters was habitat loss. Large areas of wetland across North America were drained for agricultural purposes eliminating high-quality habitats (Melquist et al. 2003). Riparian vegetation and structure for dens and cover are important components of otter habitat (Melquist and Hornocker 1983; Dubuc et al. 1990; Swimley et al. 1998). Removal of riparian habitat through development or industrial uses is a potential threat to otter populations. The Eurasian otter has gone through similar declines in Europe further demonstrating the sensitivity of river otters to a range of environmental disturbances (Foster-Turley et al. 1990). Over exploitation of otter for their fur contributed to population declines throughout North America (Foster-Turley et al. 1990; Raesly 2001). In British Columbia, otter are recognized provincially as a Class 2 (sensitive to harvest) species because their large spatial requirements necessitate management over multiple trap lines (Hatler et al. 2003). Although 4 otters contribute to the fur harvest, there is very little information to determine sustainable levels of harvest across regions and trap lines. River otters in British Columbia are found throughout the province in both marine and interior watersheds. The few studies conducted in the province have focused on river otters in coastal environments (Stenson et al. 1984; Stenson 1985; Giere and Eastman 2000). Although marine and freshwater systems have very different habitats and prey types, there have been no studies of the ecology of otters inhabiting interior fresh water systems within the province. Presently, harvest and management guidelines and knowledge of potential threats are based largely on studies conducted in Alberta and the western United States (Melquist and Hornocker 1983; Reid et al. 1994a, b). Efficient, accurate, and reliable survey methods are needed to make sound and informed management and conservation decisions. Census techniques for estimating the size, relative abundance, and distribution of otter populations are not well established. Melquist et al. (2003) suggested that the lack of a widely accepted monitoring technique for river otters is a cause for concern for wildlife managers, researchers, and conservationists. Latrine sites have the potential to provide a multitude of information concerning river otter populations. Identification, collection, and/or counting of scats at latrine sites has been used as a technique to estimate otter occupancy, distribution, habitat selection, abundance, and/or population structure (Dubuc et al. 1990; Newman and Griffin 1994; Swimley et al. 1998; Kalz et al. 2006; Prigioni et al. 2006). A thorough understanding of latrine selection and dynamics can provide important information on otter ecology that is critical in the development of conservation and management strategies. Although scat has the potential to 5 provide a wealth of information, further research is needed to develop and refine its use as a technique for monitoring otter populations. A detailed and unbiased description and explanation of the composition of diet is important for understanding the foraging ecology of otter. Past studies have simply reported descriptive statistics and graphs of prey remains (Greer 1955; Larsen 1984; Reid et al. 1994b). Scat samples are often collected during a single time period and grouped together to provide a broad assessment of otter diet. There is very little information describing the spatial and temporal variability in the diet of otters or an assessment of the predictability of these patterns. In addition, the relationship between otter diet and activity at latrine sites is not well understood. Spatial and temporal changes in the availability and use of prey items may affect the frequency of occupancy of latrine sites. Variation in activity at latrine sites will have direct implications for survey techniques that rely on the detection of otter sign (Gallant et al. 2007; LeBlanc et al. 2007). In Chapter 2, I used two different, but complimentary techniques, to increase our understanding of the composition and variation in otter diet: inventory of prey remains in scats and stable-isotope analysis of prey signatures in otter hair. I used the information from scat content analyses and scat counts to model the presence of prey items and the number of scats, respectively, in relation to variables that captured the spatio-temporal variation in the availability of prey groups. These results can help guide management strategies and refine techniques for assessing and monitoring the diet of otter populations. Studies of latrine sites in the past have typically focused on either fine- or coarsescale habitat characteristics and have not addressed the confounding effects of scale (Dubuc 6 et al. 1990; Bowyer et al. 1995; Swimley et al. 1998). Furthermore, there is very little information relating habitat characteristics at latrine sites to the consistency and intensity of activity by otters. Some latrines may have a long history of frequent use, while others may be an ephemeral site used by an otter at a single point in time. Information relating the habitat characteristics of latrines to the degree of otter activity would be beneficial to natural resource planners, managers, and foresters, wanting to incorporate the habitat requirements of otters in land-use decisions. Also, a better understanding of spatial and temporal parameters that influence otter behaviour and distribution can help improve or limit biases associated with census or survey techniques for this species. Considering this knowledge shortfall, I identified structural and vegetation characteristics of otter latrine sites at both the fine and coarse scales (Chapter 3). At the fine scale, I used habitat measurements to examine selection of latrine sites at the shoreline patch. I then used Geographic Information Systems (GIS) data to identify habitat attributes explaining the selection of latrine sites by otters at a coarser landscape scale. In addition, I used the number of otter scats recorded at latrines to investigate factors that might explain otter activity at these important habitat features. Lastly, in Chapter 4, I summarize and conclude with the findings, relevance, and management implications of my research. Study Area The research was conducted in and adjacent to the co-managed (UNBC and Tl'azt'en Nation) John Prince Research Forest (JPRF) (Figure 1). The JPRF is a 13,000-ha portion of forested crown land 45 km northwest of Fort St. James, British Columbia. The area is 7 characterized by rolling topography with low mountains (elevation range between 700 m and 1267 m) and a high density of lakes, rivers and streams. Found in the sub-boreal spruce biogeoclimatic zone, the JPRF is located between Pinchi and Tezzeron lakes and includes many smaller lakes and streams. Pinchi and Tezzeron lakes drain into the Stuart and Nechako rivers, but are not directly connected. Major drainages of Tezzeron and Pinchi Lakes are the Kuskwa and Pinchi Rivers, respectively. Major tributaries flowing into Tezzeron Lake include Grostete, Tezzeron, and Hatudetahl Creeks. The Oocock and Tsilcoh Rivers flow directly into Pinchi Lake. Tezzeron Lake's shoreline stretches for 82 km (area = 8079 ha), while the perimeter of Pinchi Lake is 67 km (area = 5586 ha) in length. The mean depth of Tezzeron and Pinchi Lakes are 11.2 and 23.9 m, respectively. Shoreline topography varies considerably along both lakes, but the area surrounding Pinchi Lake is generally more mountainous with steeper slopes. Tezzeron Lake has very little development whereas Pinchi Lake has a mercury mine (non-operational) and some residences. There is a long history of timber management and activity within the forests surrounding these lake systems. Sucker (Castomidae), trout (Salmonidae), minnow (Cyprinidae), and sculpin (Cottidae) fish families are abundant in both lake systems. Although distribution and abundance are unclear, burbot (Lota lota) are also present in both lakes. Sockeye salmon (Oncorhynchus nerka) spawn in both the Kuskwa and Pinchi Rivers during the fall. Kokanee (Oncorhynchus nerka) spawn in creeks and rivers during the late summer and early fall. A multitude of water fowl species nest on or near the two lake systems with common mergansers (Mergus merganser americanus) and red-necked grebes (Podiceps grisegena) 8 especially common on the major lakes throughout the ice-free season. Freshwater clams (Genus: Anondonta) are locally abundant in shallow areas throughout both lakes. Beaver {Castor canadensis) and muskrat {Ondatra zibethicus) are two aquatic species of mammals that are commonly documented from both sign and sightings. 9 Legend @ River otter latrine sites ——• Streams | JPRF boundary 5 km V ,\ " * Figure 1. Map of river otter study site and John Prince Research Forest (JPRF) including locations of latrine sites, 2007-2008. 10 CHAPTER 2 Spatio-temporal variation in river otter diet and latrine site activity on Tezzeron and Pinchi Lakes, British Columbia. Abstract Fluctuations in the distribution and abundance of prey resources are an important influence on the foraging ecology of carnivores. Spatio-temporal variation in the diet of river otters (Lontra canadensis), however, is not well understood. In addition, we have limited knowledge about seasonal changes in otter activity at latrine sites and how this may relate to changes in otter diet. I investigated patterns in river otter diet in central British Columbia during the ice-free season. I used a combination of scat content and stable-isotope analyses to assess the contributions of different prey items to otter diet. I used an Information Theoretic Model Comparison approach to investigate the spatio-temporal variation in the availability of prey groups as it influenced the composition of otter diet and the number of scat deposited at latrine sites. Every two weeks, I surveyed latrine sites to collect and count the number of scat deposited by otters. Scats were washed, sorted, and identified for prey remains. I used binary and count models to predict the presence of individual prey items and number of scats, respectively. For each analysis, I developed eight models using combinations of five variables representative of the spatio-temporal variation in the availability or distribution of prey items. I then compared models using Akaike's Information Criterion (AIC). I was able to develop models with good predictive power for three of the prey groups that characterized otter diet. A combination of fish spawning period, water-body type, and lake best described the presence of salmonids, minnows, and insects in otter scats. The number of scats was best described by a two-week calendar time measurement and geographic location. Scat deposition was positively influenced by a time period when no fish were spawning (early July) and to the kokanee (Oncorhynchus nerka) 12 spawning period (early September). In general, the stable-isotope analysis agreed with the results of the scat content analysis; fish dominated the diet with lesser contributions from other prey items. The stable-isotope analysis, however, suggested that sockeye salmon, large species of fish, and birds contributed more than was revealed by scat content analysis. Management decisions are often based on estimates of population trends, and this study provides the baseline information required for developing techniques for assessing and monitoring otter populations. Introduction Knowledge of the spatial and temporal variation in the utilization of food resources is important for effectively managing wide-ranging carnivores (Molsher et al. 2000; Eide et al. 2004). Seasonal variation in the use of food resources can influence the population dynamics (Fuller 1989), habitat selection (Milakovic 2008), and foraging ecology (Gonzalez 1997) of a species. Spatio-temporal variation in the diet of North American river otters (Lontra canadensis) is not well understood. Few studies have described changes in otter diet within or across seasons (Melquist and Hornocker 1983; Reid et al. 1994a), and the majority of studies have averaged diet over a broad time period with little account for seasonal differences (e.g., Gilbert and Nancekivell 1982). We have very little information to predict changes in otter diet and its associated affects on otter populations. A better understanding of the factors influencing variation in diet is critical for developing sound management and conservation strategies for river otters. 13 The majority of diet studies for otter have been premised on an analysis of prey remains in scats. These studies suggest that the diet of river otters throughout their range consists predominately of fish (Greer 1955; Larsen 1984; Reid et al. 1994b). Crayfish and frogs, however, can comprise a substantial portion of the diet of southern populations of otter (Wilson 1954; Toweill 1974). In coastal environments, both crabs and mollusks can contribute to otter diet, but occur in lower frequencies relative to fish (Larsen 1984; Stenson et al. 1984). Insects, clams, snails, snakes, turtles, waterfowl, shore birds, beaver, and muskrat have all been documented as secondary prey items for both coastal and interior populations (Toweill 1974; Larsen 1984; Stenson et al. 1984; Reid et al. 1994b). Although fish are the predominant prey item for most populations of otter, the importance of individual fish species and the occurrence of secondary prey items can vary throughout the year. In central Idaho, for example, Melquist and Hornocker (1983) found the annual frequency of occurrence of prey items in scats to be fish (97%), invertebrates (8.4%), birds (2.9%), mammals (2.6%), and reptiles (<1%). Birds reached a high of 20% in July, which coincided with the brooding and moulting periods and invertebrates reached a high of 20% in February. In contrast to this intra-year variation, mammals had a consistent and low frequency of occurrence throughout the year. Among fish species, kokanee salmon (Oncorhynchus nerka), mountain whitefish (Prosopium williamsoni), mottled sculpin (Cottus bairdi), and large-scale suckers (Catostomus macrocheilus) were the four species dominating otter diet. Large-scale suckers were recorded most frequently in spring and summer when they enter streams to spawn, and kokanee salmon were recorded most frequently in the fall during their spawning period. Mountain Whitefish were recorded most frequently in the 14 winter, while mottled sculpin occurred in 31-42% of scat during each season throughout the entire year (Melquist and Hornocker 1983). In the boreal forests of northeastern Alberta, Reid et al. (1994b) also found fish to be most abundant in otter scats with other prey contributing less but displaying distinct seasonal variations. Birds in the diet peaked at a 15% frequency of occurrence in July and insects peaked at approximately 73% in August. Reid et al (1994b) determined that mollusks were an important part of otter diet with a peak frequency of occurrence of 25% in September. Among fish species, white sucker (Catostomus commersoni) and 4 species of fishes in the Cyprinidae family were the most dominate components of otter diet. Cyprinids and gasterosteids were dominant in the winter months, catostomids were dominant in open-water months, and salmonidae were dominant in October. Crait and Ben-David (2006) also found a distinct seasonal peak of 79% occurrence of a salmonidae species during the spawning period. Although scat inventories can index the relative importance of different food categories, this technique has several known biases (Larsen 1984). First, the method relies on hard parts found in the scats that can then be identified to species or family. Differences in the number, size, and digestibility of hard parts across individual prey species have the potential to bias diet estimates. Second, a high frequency of occurrence of hard parts in scats does not directly relate to the biomass consumed by otters. For example, Reid et al. (1994b) reported that the high frequency of insect remains in scats during the summer season may not be a meaningful indicator of the total biomass consumed. Finally, scats provide information only on an animal's recent meals and large numbers of scat through time may be needed to 15 draw accurate and unbiased conclusions about seasonal or annual patterns in prey consumption. Despite these biases, Erlinge (1968) concluded through captive feeding trials that frequency of occurrence data gave a reasonably accurate measure of the relative importance of prey species. Additional methods for measuring diet, however, would be useful as a complimentary technique for revealing these biases and producing the most accurate representation of otter diet. As an alternative or complimentary technique, stable isotopes are being used with increasing frequency to investigate animal diets (Urton and Hobson 2005; Mowat and Heard 2006; Milakovic 2008). This technique measures the ratios of naturally occurring carbon and nitrogen stable isotopes in blood, tissue, bone, or hair and relates that information to the isotopic signatures of potential prey items (Dalerum and Angerbjorn 2005). Hair has been used with stable-isotope analyses to investigate the composition of diet during the time periods of hair growth (Dalerum and Angerbjorn 2005). The metabolically inactive portion of the hair will retain isotopic signatures characteristic of the diet during the period of hair growth. In addition, many animals experience an annual moult that allows for predictable seasonal growth periods in hair. Isotopic analysis of hair has been used to investigate seasonal shifts in the diets of wolves (Canis lupus; Darimont and Reimchen 2002), brown bear (Ursus arctos; Ben-David et al. 2004), and arctic fox (Alopex lagopus; Roth 2002). North American river otter have two cycles of hair shedding and replacement during the year. From May through August, otters shed and replace under fur, and from August to November they shed and replace guard hair (Ben-David et al. 2000). This pattern of hair loss and growth appears common among carnivores inhabiting northern environments and when 16 combined with stable-isotope analysis allows for the identification of bi-seasonal patterns in diet (Harper and Jenkins 1982; Maurel et al. 1986). Past studies of otter diet have reported descriptive statistics and graphs of prey remains throughout the year (Greer 1955; Larsen 1984; Reid et al. 1994b). No attempts have been made to model otter diet through space and time or to assess the predictability of these patterns using information from scat inventories. In addition, stable-isotope analysis of otter hair and their prey sources has not been conducted on otter populations in freshwater systems. The use of stable isotopes has the potential to aid in the interpretation of biases associated with fecal inventory analysis. Stable isotopes could provide an easier and more cost effective means of monitoring otter diet over large geographic areas. Lastly, there is little information relating changes in the food consumed by otters to fluctuations in scat deposition rates. No attempts have been made to model spatio-temporal variation in the number of scats deposited by otters at latrine sites. Changes in otter diet through time may affect scat deposition rates and the location of latrines. Such variation will have implications for population or habitat surveys that rely on otter sign. Otter are assumed to deposit scats in areas where they are feeding, and increased activity at latrine sites may be associated with abundant food resources. The food items, time period, and habitat types associated with these high-activity areas may be especially important in the ecology and life history of otter populations. Surveys that rely on otter sign to estimate otter distribution or abundance will produce biased results if spatial and temporal variability is not taken into account. 17 I used an Information Theoretic Model Comparison approach to investigate variation in the diet of river otter and deposition rates at latrine sites in central British Columbia during the ice-free season. I used a combination of prey remains in scat and stable isotopic signatures in otter hair to assess the contributions of different prey items to otter diet. I measured deposition rates at known latrines and used those data as a measure of otter behaviour. Both sets of data were related to explanatory variables that indexed the spatiotemporal variation in the availability of prey items. My specific objectives were to: (1) document the diet of otter in an interior freshwater system of British Columbia during the ice-free season using a combination of assessment techniques; and (2) investigate the relationship among otter diet, spatio-temporal variation in resources, and scat deposition rates. Methods Data Collection Locations of latrine sites were identified by a series of shoreline surveys on both Pinchi and Tezzeron Lakes as well as along all significant tributary streams (1 km from lakestream confluence and navigable by canoe or kayak). I conducted shoreline surveys by canoe, kayak, and on foot. Two complete surveys of all shorelines were conducted in 2007. The first occurred from 5-27 June and the second from 20 July to 5 August. In 2008, I randomly selected and intensively surveyed 200, 200-m segments of shoreline split evenly between the Tezzeron and Pinchi lake systems from 15 August to 15 September. The 2008 survey was conducted to both identify new latrine sites and to determine if I found most or 18 all of the active latrine sites in 2007. I surveyed latrine sites every two weeks to collect scats and record the frequency of use. In 2007, scat collection began in July and ended in late October. In 2008, scat collection began in mid-May and ended in mid-October. I randomly selected a sub-sample of scats at latrine sites for collection and diet analysis. I attempted to leave -25% of the scat at latrine sites to maintain active scent for revisits by otters. Scats not collected at latrine sites were marked with silver glitter to differentiate old from new samples during surveys. Continued high activity at latrine sites during the duration of the study suggested my surveys had minimal effect on otter visitation rates. Scats were stored in Ziploc bags, frozen after collection, and later washed through a lxl-mm strainer and airdried. Scats were sub-sampled for prey identification using a stratified random sampling method to maintain relatively equal sample sizes among time periods and locations. After washing and drying, the entire contents of each scat were spread evenly over a grid containing 36 lxl-cm cells. I randomly selected eight of the 36 cells for prey identification. Each cell was marked for its contents: bone, scale, hair, feather, clam shell, and/or insect exoskeleton. Scales were used as the primary identifier of fish family groups and supported by intact jaw bones with teeth. Identification of fish remains was conducted using a reference collection that I generated for the study area as well as available literature (McAllister and Lindsey 1961; Lagler 1970; Nelson 1973; Cannon 1987). I used a 10-20x dissecting microscope to identify scales. A total of five fish family groups were identified in the scat content analysis (Table 1). I was able to also identify the two subfamilies of Salmonidae. Mammals and birds were primarily identified by the presence of hair and 19 feathers, respectively. Insects were identified by exoskeletons, and mollusks by shell fragments. Composition of food types within scat was summarized as the percentage of the total number of scats analysed that contained a particular prey type. Prey tissue samples for stable-isotope analysis were collected during late summer of 2008 (British Columbia Ministry of Environment Fish Collection Permit No: PG08-45338). Potential prey items were assessed from a review of relevant literature on otter diet (Toweill 1974; Melquist and Hornocker 1983; Larsen 1984; Reid et al. 1994b), from published accounts of fish and wildlife species in the study area (Dodds 2001; McPhail 2007), and from initial observations of prey remains in otter scat. I attempted to collect consumable tissue (i.e., muscle) from three different individuals within each prey group. Prey tissue samples were dried at approximately 60-70° C for 48 hrs and ground to a fine powder using a WIG-LBUG grinder (Crescent Dental Company, Chicago, Illinois, USA) (Ben-David et al. 1997a). Guard hairs were pulled from river otter during capture efforts associated with a radiotelemetry study on Tezzeron and Pinchi Lakes in October 2007 and May 2008. In addition, guard hair samples were collected from otters commercially trapped in the vicinity of the study area (~ 30 km radius) during November 2008. Hair was cleaned of surface oils in 2:1 chloroform:methanol solution, rinsed with distilled water, and air-dried (Hobson et al. 2000). A subsample (1-1.2 mg) of the fine powder tissue sample or intact hairs were weighed into a tin capsule (5x9 mm) (Costech Analytical Technologies Inc., Valenica, CA, USA), encapsulated, and placed into a 96-well plate. Samples were measured for 513C and 515N isotopes at the University of California Davis Stable Isotope Facility using a continuous flow isotope-ratio mass spectrometer. 20 Table 1. Fish family groups identified in river otter scats on Tezzeron and Pinchi Lakes, central British Columbia during the ice-free season (2007-2008). Family Subfamily Salmonidae Salmoninae Common Name Salmon Salmonidae Coregoninae Whitefish Cyprinidae Minnow Cottidae Sculpins Castomidae Sucker Gadidae Cod Potential species included Sockeye salmon {Oncorhynchus nerka) Lake trout {Salvelinus namaycush) Rainbow trout {Oncorhynchus mykiss) Kokanee {Oncorhynchus nerka) Lake whitefish {Coregonus clupeaformis) Mountain whitefish {Prosopium williamsoni) Lake chub {Couesius plumbeus) Redside shiner {Richardsonius balteatus) Longnose dace {Rhinichthys cataractae) Northern pikeminnow {Ptychocheilus oregonensis) Prickly sculpin {Cottus asper) Slimy sculpin {Cottus cognatus) Longnose sucker {Catostomus catostomus) White sucker {Catostomus commersonii) Largescale sucker {Catostomus macrocheilus) Burbot {Lota lota) 21 Stable-isotope Analysis For the stable-isotope analysis, I used Isosource (Version 1.3.1; Phillips and Gregg 2003), a dietary mixing model, to quantify the relative range in proportions of potential prey items within otter hair (Phillips and Gregg 2003; Phillips et al. 2005). In this model, all combinations of each source contribution are examined and any combination that sums to the observed mixture of isotopic signatures is a feasible solution (Phillips and Gregg 2003). For this reason, a distribution of feasible solutions is reported rather than a single value. The model requires that isotopic signatures be corrected for enrichment from prey to consumer. Consumer diet-to-hair fractionation values are not established for river otters; however, I used a fractionation value of 1 0/0° for Carbon and 2 0/0° for nitrogen based on captive feeding trials of mink and bear (Ben-David 1996; Hilderbrand et al. 1996). These values were used for two reasons: 1) they are enrichment values based on the most functionally related carnivore species available; and 2) these values have been used with success on a study of river otters in a coastal environment (Ben-David et al. 1998). I investigated the sensitivity of the analysis to variations in carbon and nitrogen fractionation values by examining prey source contributions when values were manipulated up to +15%. I set a tolerance of 0.10/0° and source increment value of 1% to incorporate measurement error and sample variability within the Isosource model (Urton and Hobson 2005). The most important factor in controlling uncertainty in estimates of source proportions is the existence of large isotopic differences among sources (Phillips and Gregg 2001). Due to a large number of potential sources, I used a combination of a priori and posteriori methods to combine sources for mixing analysis (Phillips et al. 2005). I first 22 combined sources a priori into logically related groups based on the biology (e.g., trophic level, marine vs freshwater, terrestrial vs aquatic) of prey sources and on the relative similarity of isotopic signatures. When the Isosource model produced indeterminate results (broad range of solutions), I examined the ranges and aggregated the sources where needed. Sources were grouped only if the majority of their estimated contributions (>75%) overlapped and ecological significance (i.e., functionally related prey) was maintained. The combination of methods resulted in narrower and constrained results for interpretation. Diet and Scat Model Development For each prey group identified in otter diet, I used logistic regression to identify factors that explained variation in occurrence within sampled scats (Menard 2001). The presence (1) or absence (0) of a particular prey group was the dependent variable in the model. I constructed a zero-inflated negative binomial (ZINB) count model to investigate the relationship between the number of scats deposited at a latrine site and variables that represented the spatio-temporal variation in the availability of prey groups (Nielsen et al. 2005). A negative binomial regression allows for the over dispersion of counts that is often characteristic of ecological data (Long and Freese 2006). I tested this assumption with a likelihood ratio test. A zero-inflated model incorporates excess zero counts through a mixture of two separate processes: a data generating process of zeros and another of either a Poisson or negative binomial distribution (Long and Freese 2006). I used a Vuong test to determine if a zero-inflated model was appropriate (Vuong 1989). All data analyses were performed using Stata (ver.9.2, Statacorp, 2006). 23 I used five spatio-temporal variables to explain both the presence of diet items in individual scat samples (logistic regression) and the frequency of scats (zero-inflated negative binomial) at latrine sites (Table 2). Variables representing the spatial distribution of latrine sites included; geographic zone (classified by the adjacency (<1 km) and location of sites along the shoreline), and water-body type (stream, open lake, reedy bay). I hypothesized that the distribution and abundance of the range of prey species varied spatially throughout the two lake systems. Thus, geographic zones were used to represent five distinct areas across each lake. Drawing on my understanding of fish ecology, I hypothesized that there would be spatial heterogeneity in the diet of otters because of variation in the life history requirements of major fish species. Different prey species and cohorts inhabit different water-body types throughout the year depending on their habitat requirements. For example, rainbow trout are found in both stream and lake habitats, while lake trout are found almost exclusively in deep water lakes (McPhail 2007). The movement of kokanee from lake into stream habitats to spawn during the fall is another example of a variable spatial distribution related to habitat (McPhail 2007). Temporal variables included season (measured as sequential two-week periods throughout the study's duration), and fish spawning periods (time of spawning for fish species found in the study lakes). The season variable captured the variability in the distribution and abundance of all potential prey species throughout the year. For example, burbot move from deep water in the summer to shallow water in the fall in response to changing temperatures (McPhail 2007). 24 Table 2. Parameters used in the development of binary models for presence of river otter prey items and for ZINB count models of scat numbers on Tezzeron and Pinchi Lakes in central British Columbia, from 2007-2008. The number of parameters or categories within each variable are represented by K. Parameter Description season seasonal increments in 2-week periods. geozone geographic zone (separated by latrine site clusters and distance) waterbody water-body type (stream, open lake, reedy bay) fishspawn occurrence of fish spawning period lake latrine occurred on Tezzeron or Pinchi Lake 25 Variable type (K) categorical (9) categorical (10) categorical (3) categorical (4) categorical (2) For the fish spawning variable, I hypothesized that fish would be most vulnerable to otter predation when they aggregate in shallow water to spawn and, thus, should constitute a larger portion of otter diet during those periods. The timing of spawning for major fish groups was estimated from published data on life history strategies and local knowledge of fish in the study area (McPhail 2007). This resulted in four time periods: late June - early August (fishspawn 1 = no spawn), early June (fishspawn 2 = rainbow trout, sucker, lake chub, northern pikeminnow, sculpin), late September - October (fishspawn 3 = sockeye salmon, lake trout, whitefish), and late August - early September (fishspawn 4 = kokanee). As a fifth variable, I modeled the effect of the larger geographic areas, Tezzeron and Pinchi Lake, as a possible explanatory factor for variation in otter diet and scat deposition rates. Variations in the topography, water depths, and habitat characteristics of the two lakes may affect the relative distributions and abundance of different prey sources. I used biologically plausible combinations of the five temporal and spatial covariates to develop each explanatory model. I combined the spatial and temporal variables of geozone and season into the same model because they were measurements of space and time not directly correlated with fish prey sources. Water-body type and spawning period were included together because they were spatio-temporal measurements based on descriptions of fish prey sources. I used eight models as hypotheses to explain the presence of each food group and the number of scats at latrine sites (Table 3). Year was not included in any of the models due to differences in the start and end date of data collection between years. I used deviation coding (desmat.ado; Hendrickx 2001) to represent all categorical variables and variance inflation factors (VIP) to assess multicollinearity. 26 Table 3. A priori candidate models for the binary (prey presence) and count analyses (number of scats) of river otter diet and scat deposition on Tezzeron and Pinchi Lakes in central British Columbia, based on data collected from 2007-2008. The number of parameters or categories within each variable are represented by K. Model covariates geozone+season season geozone waterbody+fishspawn waterbody fishspawn geozone+season+lake waterbody+fishspawn+ lake Rationale Assess role of season and geographic area on presence of prey items and scat counts, Assess role of season on presence of prey items and scat counts, Assess role of geographic area on presence of prey items and scat counts, Assess role of water-body type and fish spawning period on presence of prey items and scat counts, Assess role of water-body type on presence of prey items and scat counts Assess role of fish spawning period on presence of prey items and scat counts, Assess role of season, geographic area, and lake on presence of prey items and scat counts. Assess role of water-body type, fish spawning period, and lake on presence of prey and scat counts. 27 K^ 18 9 10 6 3 4 19 7 An individual VIF value >10 or a mean VIF value >1 suggested that a model had high levels of multicollinearity (Chatterjee et al. 2000). Diet and Scat Model Selection I used Akaike's Information Criterion (AICC) for small sample sizes to identify the most parsimonious explanatory models of otter diet and scat numbers (Burnham and Anderson 2004). The AICC values are a relative metric that must be compared in the context of a set of a priori models (Table 3). I used both AAICC and Akaike weights (AICvv) to rank and compare models. The model with the lowest AICC score is considered the "best" or the most parsimonious model given the data and the set of models compared. A model with a AAICC <2, however, was considered to be equivalent to the model with the minimum score (Burnham and Anderson 2002). When models had AAICC values that were nearly equivalent, I selected the most parsimonious model (i.e., fewest number of parameters). An AICvv is a value from 0-1 that represents the approximate probability that a model is the best among a set of candidate models. I used beta-coefficients and z-statistics (P < 0.05) to assess the importance of model parameters. Given the large set of models that I tested and for ease of interpretation, I present coefficients from only the most parsimonious model. Predictive Ability of Diet and Scat Models Data for all count models were randomly divided into training (85%) and testing (15%) groups using a random number generator and a uniform distribution (Fielding and Belll997). I used the receiver operating characteristics (ROC) and resulting area under the 28 curve (AUC) to assess the predictive ability of the "best" model from the binary analyses. The AUC measures the relative proportions of correctly and incorrectly classified predictions (Pearce and Ferrier 2000). AUC values 0.5 to 0.7 were considered to have poor model accuracy, from 0.7 to 0.9 good model accuracy, and >0.9 were considered to have high model accuracy (Swets 1988). I used Pearson's standardized residuals to identify outliers (Menard 2001). I used the predicted counts as well as the predicted probabilities of counts to evaluate the performance of the most parsimonious count models (prcounts.ado; Long and Freese 2006). I evaluated the performance of the model by visual inspection of graphs plotting the observed probability of a count using the model testing data and the predicted probability of a count generated from model training data. The residual difference between observed and predicted counts allowed me to further examine the models predictive ability across the range of values I observed. Results Scat content analysis I counted 4,470 scats at latrine sites during the duration of the study. A sub-sample of 901 scats were cleaned and sorted for hard part identification. The diet of otter was comprised mostly of fish, with sucker species being the most dominant (Figure 2). Throughout the icefree season, the percentage of sucker species remained relatively constant and high, sculpin species remained constant and low and the remaining species were intermediary with fluctuations in their frequency of occurrence throughout the season (Figure 3). 29 60.0% « 50.0% g- 40.0% £ 30.0% a> o 20.0% -\ a. Prey Group Figure 2. Percent frequency of occurrence of prey groups in river otter scats on and near Tezzeron and Pinchi Lakes, central British Columbia from late May to October, 2007-2008. 30 60.0% Season Figure 3. Percent frequency of occurrence of fish prey groups in river otter scats on and near Tezzeron and Pinchi Lakes, central British Columbia from late May to October, 2007-2008 (E = early, L = Late). 31 9.0% 8.0% -#-MAMMAL 7.0% -i 3-BIRD c 6.0% CLAM g- 5.0% -\ -X-INSECT ~ 4.0% | 3 0 % °" 2.0% 1.0% ^ 0.0% Season Figure 4. Percent frequency of occurrence of non-fish prey groups in river otter scats on and near Tezzeron and Pinchi Lakes, central British Columbia from late May to October, 20072008 (E = early, L = Late). 32 The only large decrease in sucker occurrence was in late August and corresponded with a peak in the frequency of salmonid species. Although the occurrence of non-fish prey items fluctuated throughout the season, their frequency was consistently low (Figure 4). Stable Isotopes A total of 22 otter hair samples were combined for the stable-isotope analysis (Appendix 1). Values for insects (mayfly, stonefly, caddisfly) were taken from a study in the nearby Stuart River watershed (Johnston et al. 1997); I used values for insects collected in areas with low sockeye salmon density. I adopted a two-step process to overcome the low resolution associated with increased complexity (i.e., too many sources) in the Isosource model. First, I combined the majority of fish species into a single group for comparison with all other food items based on the hypothesis that fish would comprise a large proportion of the diet. In the first model, each of the estimated contributions of the non-fish food sources overlapped zero; thus, non-fish species were removed from the second model to examine the relative contributions of different fish groups. For the first model, the combination of a priori and posteriori methods for prey aggregation resulted in six potential prey sources based on their ecological, taxonomic, and relative isotopic similarities (Figure 5). First, all fish werecombined into a single prey source with the exception of sockeye salmon (Figure 5, Fish 1). Sockeye salmon are a marine-derived food source and differ greatly in their ecology and isotopic signatures from other freshwater inhabiting species (Ben-David et al. 1997a). The fish species included in the aggregate group were rainbow trout, lake trout, kokanee, sucker species, northern pikeminnow, whitefish, burbot, and sculpin species (Fish 1). 33 The remaining four prey sources belonged to different taxonomic groups and had dissimilar isotopic values (mammal, bird, clam, and insect) (Figure 5). The stable-isotope model produced results with a narrow range of solutions that were easily interpretable. The aggregate fish group (Fish 1) was the dominate prey contributing to otter diet during the hair growth period, followed by sockeye salmon (Model 2, Table 4). Bird species were also a contributing source to otter diet, but had a wider range of potential solutions. The stable-isotope data suggested that all other prey sources had a very small contribution. From the results of the first mixing analysis, I determined that fish were the dominant food source for otter and that non-fish prey items contributed considerably less. All of the non-fish prey sources overlapped zero and were removed for the second stable-isotope model to examine the relative contributions of different prey groups. Based on the results of the first mixing model, I divided the fish into separate groups based on their ecology and relative isotopic similarity for the second model: 1) kokanee (unique ecology and isotopic signatures), 2) sockeye salmon (marine-derived and unique isotopic signatures), 3) burbot and lake trout (large deep-water carnivorous fish), and 4) rainbow trout, sucker species, northern pikeminnow, whitefish, and sculpin species (very similar isotopic signatures) (Fish 2, Figure 5). Once again, the aggregate fish group (Fish 2) was the dominate food source (Model 2, Table 4). Both the sockeye salmon and the burbot/lake trout prey source contributed equally to the mixing model describing the diet of otter. Kokanee was a very small contributing source to the solution. 34 14 Lake Trout/Burbot Salmon Kokanee 10 Fish ^ H ^ f -32 -30 lo 515 N 6H i Clam. t e r Fish 2* Bird * -34 12 Insect 1 Mammal Mar -28 -26 -24 2 -22 -20 13 5'°C Figure 5. Mean (+ SE) stable isotopic signatures for potential prey items of river otters on Tezzeron and Pinchi Lakes, central British Columbia. *Fish 1: rainbow trout, lake trout, kokanee, sucker species, northern pikeminnow, whitefish, burbot, and sculpin species. **Fish 2: rainbow trout, sucker species, northern pikeminnow, whitefish, and sculpin species. 35 Table 4. Estimated range (%) of prey sources in river otter diet on Tezzeron and Pinchi Lakes, central British Colubmbia, from 2007-2008. Estimates generated by analysis of stable isotopes using a dietary mixing model (Isosource), including a sensitivity analysis with a 15% change in fractionation values for carbon (AC) and nitrogen (AN). Model 1: All Prey Items Salmon Fish 1* Clam Bird Mammal Insect Range 14.0 - 22.0 50.0-80.0 0.0-5.0 0.0 - 27.0 0.0 - 6.0 0.0- -4.0 Range AC 11.0-28.0 31.0-86.0 0.0-8.0 0.0-41.0 0.0- -4.0 0.0 - 8.0 2.0-57.0 Range AN 0.0 - 28.0 14.0-80.0 0.0-13.0 0.0- -7.0 0.0-12.0 Range AC+AN 9.0 - 32.0 4.0-76.0 0 . 0 - 1 5 . 0 0 . 0 - 6 3 . 0 0.0- -8.0 0 . 0 - 1 4 . 0 Model 2: Fish Prey Items Salmon Fish 2** Burbot/Lake trout Kokanee Range 10.0- 17.0 66.0 - 82.0 7.0 •• 17.0 1.0- 10.0 Range AC 10.0 - 20.0 65.0 - 70.0 4.0 - 16.0 3.0- 12.0 Range AN 10.0- 17.0 66.0-76.0 5.0- 17.0 0.0- 10.0 Range AC+ AN 6.0 - 25.0 52.0 - 80.0 2.0 -•18.0 0.0- 20.0 *Fish 1: rainbow trout, lake trout, kokanee, sucker species, northern pikeminnow, whitefish, burbot, and sculpin species. **Fish 2: rainbow trout, sucker species, northern pikeminnow, whitefish, and sculpin species. 36 Although variation was noted and some changes produced a wide range of solutions, the overall pattern of results was insensitive to small changes in fractionation values (Table 4). Diet Models For the logistic regression analysis of the presence-absence of diet items in scats, I was not able to construct models for several of the prey groups because of very low frequencies. These prey groups included mammal, bird, clam, and burbot. I was able to develop a set of models for five different fish groups: minnows, salmonids, suckers, sculpins, and whitefish. I was also able to model the presence/absence of insects. None of the model variables had excessive multicollinearity (i.e., VIF>10). The water-body type model, followed by the water-body type and spawning period model provided the greatest support for the occurrence of sucker in scats, but both models had low predictive power (AUC = 0.574 and 0.594, respectively). For the sculpin family, three models were essentially equivalent: waterbody+fishspawn+lake, waterbody+fishspawn, and waterbody. All three models, however, had poor model accuracy using model testing data (AUC < 0.500). The whitefish family also had three models that were considered equivalent: waterbody+fishspawn+lake, geozone+season, and geozone+season+lake. All of these models, however, had poor model accuracy for model training data sets (AUC = 0.683, 0.653, and 0.642, respectively). Only three of the prey groups resulted in models with good to excellent predictive ability. 37 Table 5. Summary of AICC model selection statistics for candidate binary models used to predict prey occurrence in river otter scats on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008. Results provided only for prey group models with good predictive performance. Salmonid Binary Model waterbody+fishspawn+lake geozone+season waterbody+fishspawn geozone+season+lake geozone waterbody season fishspawn Minnow Binary Model waterbody+fishspawn+lake geozone+season season geozone+season+lake waterbody+fishspawn fishspawn geozone waterbody Insect Binary Model waterbody+fishspawn+lake waterbody+fishspawn geozone+season+lake geozone+season fishspawn season waterbody geozone Rank 1 2 3 4 5 6 7 8 AICC 408.6 422.9 422.9 425.0 432.8 438.6 443.7 446.5 AIC C A 0.0 14.3 14.3 16.4 24.2 30.0 35.1 37.9 AICw 0.998 0.001 0.001 <0.001 <0.001 <0.001 <0.001 <0.001 1 2 3 4 5 6 7 8 545.4 546.7 546.9 548.8 561.7 564.9 578.2 591.0 0.0 1.3 1.5 3.4 16.3 19.5 32.8 45.5 0.455 0.242 0.218 0.085 <0.001 <0.001 <0.001 <0.001 1 2 3 4 5 6 7 8 243.1 251.9 253.8 254.7 258.0 261.1 263.1 268.5 0.0 8.8 10.6 11.5 14.8 17.9 20.0 25.4 0.980 0.012 0.005 0.003 0.001 <0.001 <0.001 <0.001 38 The same model best explained the presence of salmonid, minnow, and insect groups and included the combination of spawning period, water-body type, and lake (Table 5). For the salmonid group, the Akaike's weight indicated the top-ranked model had a 99% chance of being the best model. The model had good predictive ability for both the model training (AUC = 0.738) and model testing (AUC = 0.808) data groups. Several coefficients were significant for the salmonid model. The presence of salmonids in otter scats was positively influenced by stream water-body types, a late August-early September spawning period, and Pinchi Lake (Table 6). Statistically significant negative coefficients included lake waterbody type, late June-early August spawning period, and Tezzeron Lake. Three models explaining the presence of minnow in otter scats had nearly equivalent AICC scores (Table 5). The highest ranked model (waterbody+fishspawn+lake) also had the fewest number of parameters (K = 7) and, thus, was the most parsimonious and the best of the set. Given the evidence of a clear best model in most cases and for ease of interpretation, I only examined coefficients for the most parsimonious model. Other spatio-temporal variables (geozone and season) from the second and third ranked model may also have influenced the presence of minnow in otter scat. The AUC demonstrated that the waterbody+fishspawn+lake model had good predictive ability for both the model training and model testing data sets with scores of 0.732 and 0.850, respectively. Coefficients indicated that the late June-early August spawning period and Pinchi Lake had a negative influence on the presence of minnow, while the late September-October spawning period and Tezzeron Lake had a positive influence (Table 6). 39 Table 6. Estimated coefficients for AICC selected models (binary) describing the presence of prey items in river otter (Lontra canadensis) diet on Tezzeron and Pinchi Lakes, central British Columbia (2007-2008). Parameter Salmonid (Salmonidae) waterbody 1 (lake) waterbody 2 (reedy bay) waterbody 3 (stream) fishspawn 1 (L June - E Aug) fishspawn 2 (E June) fishspawn 3 (L Sept - Oct) fishspawn 4 (L Aug - E Sep) lake 1 (Tezzeron) lake 2 (Pinchi) Constant Minnow (Cyprinidae) waterbody 1 (lake) waterbody 2 (reedy bay) waterbody 3 (stream) fishspawn 1 (L June - E Aug) fishspawn 2 (E June) fishspawn 3 (L Sept - Oct) fishspawn 4 (L Aug - E Sep) lake 1 (Tezzeron) lake 2 (Pinchi) Constant Insect waterbody 1 (lake) waterbody 2 (reedy bay) waterbody 3 (stream) fishspawn 1 (L June - E Aug) fishspawn 2 (E June) fishspawn 3 (L Sept - Oct) fishspawn 4 (L Aug - E Sep) lake 1 (Tezzeron) lake 2 (Pinchi) Constant Coef. SE 95% CI P -0.812 -0.069 0.881 -0.932 0.100 0.062 0.770 -0.555 0.555 -2.488 0.233 0.252 0.187 0.247 0.328 0.236 0.217 0.141 0.141 0.171 - 1 . 2 6 8 - -0.355 -0.563 -- 0.425 0 . 5 1 4 - -1.248 -1.416 —- -0.448 -0.543 -- 0.743 -0.401 --0.525 0.345 --1.195 -0.831 -- -0.279 0.279 --0.831 -2.823 ---2.152 <0.001 0.785 <0.001 <0.001 0.760 0.792 <0.001 <0.001 <0.001 <0.001 -0.298 0.235 0.064 -0.645 0.293 0.736 -0.383 0.541 -0.541 -2.078 0.163 0.166 0.155 0.202 0.247 0.172 0.237 0.135 0.135 0.145 -0.617 —-0.0215 -0.090 --0.560 -0.240 --0.368 -1.041 —- -0.249 -0.190 —-10.777 0.399 --1.073 -0.848 -- 0.082 0.277 -- 0.805 -.0806 -- -0.277 -2.361 ---1.793 0.067 0.157 0.682 0.001 0.235 <0.001 0.106 <0.001 <0.001 <0.001 -0.675 -0.043 0.718 0.654 1.278 -0.912 -1.020 0.493 -0.493 -3.70 0.321 0.311 0.256 0.310 0.377 0.487 0.568 0.215 0.215 0.302 -1.304 —- -0.046 -0.653 -- 0.567 0 . 2 1 6 - -1.220 0.046 --1.262 0.539 --2.017 -1.867 —- 0.0425 - 2 . 1 3 3 - - 0.093 0.072 --0.914 - 0 . 9 1 4 - - -0.072 -4.292 ---3.100 0.035 0.889 0.005 0.035 0.001 0.061 0.073 0.022 0.022 <0.001 40 Table 7. Summary of AICC model selection statistics for candidate ZINB models of counts of river otter scats using spatio-temporal variables. Data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008. Model geozone+season geozone+season+lake geozone waterbody+fishspawn+lake waterbody+fishspawn fishspawn season waterbody Rank 1 2 3 4 5 6 7 8 AICC 4373.7 4374.7 4382.2 4382.3 4385.5 4389.7 4393.1 4403.6 41 AICCA 0.0 1.0 8.5 8.6 11.8 16.0 19.4 29.9 AIO 0.607 0.374 0.009 0.008 0.002 <0.001 <0.001 <0.001 The top-ranked model for the insect prey group, spawning period+water-body type+lake had a large Akaike's weight suggesting that this was the best model of the set (98%, Table 5). This model had good predictive ability for both the model training (AUC =0.790) and model testing (AUC = 0.741) data sets. Lake water-body type and Pinchi Lake had a negative influence on the presence of insect remains in the otter scats that I analysed. Alternatively, stream water-body type, early June and late June-early August time periods, and Tezzeron Lake had a positive influence on the presence of insects in otter scats (Table 6). Scat Deposition Models The number of otter scats at a latrine site was best explained by a model that contained covariates for season and geographic zone (Table 7). The second ranked model consisting of season, geographic zone, and lake had an AICC score only 1.0 point higher than the top-ranked model. The second ranked model was the same as the top-ranked model with one additional variable. In this case, I selected the top-ranked model (season, geographic zone) because the additional variable lake did not compensate for the resulting loss in parsimony (Burnham and Anderson 2002). The large geographic area of lake may have also influenced the number of scat. Two areas on Tezzeron Lake (Big Bay and North side) had a positive influence on the number of otter scats, while two areas on Pinchi Lake (Southeast and West Bay) had a negative influence (Table 8). The number of scats was positively associated with a time period in early July and early September. 42 Table 8. Estimated coefficients for the single model (ZINB count) with the lowest AICC score describing the effects of spatio-temporal variables on intensity of activity at latrine sites by river otter. Data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008. Parameter Count Portion geozone tezzSE tezzmid tezznorth tezzbig bay tezzwest pinSE pinNE pinsouth pinmid pinwestbay season early June late June early July late July early August late August early September late September early October constant Coef. SE 95% CI P -0.205 0.036 0.768 0.409 0.071 -0.385 -0.152 -0.027 0.068 -0.583 0.110 0.133 0.293 0.152 0.124 0.113 0.120 0.196 0.133 0.170 -0.421—0.011 -0.225 — 0.297 0.194—1.342 0.111—0.707 -0.172 — 0.314 -0.606 —-0.164 -0.387 — 0.083 -0.411—0.357 -0.193 — 0.329 -0.916 —-0.250 0.121 0.823 0.011 0.021 0.528 0.001 0.386 0.766 0.537 0.002 -0.226 -0.253 0.367 0.036 -0.066 0.048 0.163 -0.146 0.076 1.898 0.123 0.148 0.126 0.125 0.115 0.116 0.133 0.117 0.133 0.058 -0.467 — 0.015 -0.543 — 0.037 0.120 — 0.614 -0.209 — 0.281 -0.291—0.159 -0.179 — 0.275 -0.098 — 0.424 -0.375 — 0.083 -0.185 — 0.337 1.784 — 2.012 0.177 0.045 0.004 0.911 0.619 0.568 0.044 0.083 0.666 <0.001 43 Table 8. Continued Parameter Coef. Inflate Portion geozone tezzSE 0.634 tezzmid 0.366 tezznorth -0.420 tezzbig bay -1.282 tezzwest 0.255 pinSE 0.024 pinNE 0.473 pinsouth -0.497 pinmid 0.232 pinwestbay 0.218 season early June 0.095 late June -1.741 early July -0.291 late July -0.621 early August 0.310 late August 0.442 early September 1.186 late September -0.020 early October 0.639 constant -1.670 SE 95% CI P 0.268 0.330 0.902 0.860 0.320 0.348 0.305 0.654 0.353 0.505 -0.108 — 1.159 -0.280—1.012 -2.188 — 1.346 -2.968 — 0.402 -0.372 — 0.882 -0.658 — -0.706 -0.125 — 1.070 -1.778 — 0.784 -0.460 — 0.923 -0.771 —-1.207 0.018 0.267 0.641 0.136 0.426 0.945 0.121 0.447 0.511 0.666 0.374 1.530 0.420 0.559 0.345 0.332 0.311 0.395 0.339 0.289 -0.638 — 0.828 -4.739 — 1.257 -1.114 — 0.533 -1.717 — 0.474 -0.366 — 0.985 -0.210—1.093 0.576—1.796 -0.795 — 0.754 -0.025 — 1.303 -2.236—-1.104 0.799 0.255 0.489 0.266 0.369 0.184 <0.001 0.959 0.059 <0.001 44 0.4 -Observed Probability of Count (85%data) 0.35 Predicted Probability of Count (15%data) 0.3 +•» c § 0.25 *^ o & 0.2 j§ 0.15 o w Q. 0.1 0.05 L:-*~.,<_ 1 0 4 1 r 6 8 10 12 '"*™*-"R=HB—•- 14 16 18 f^r'V'W-^-A'-T'VH 20 22 24 26 28 30 Scat count Figure 6. Comparison of the best ZINB model predictions versus observed probabilities of counts of river otter scats at latrine sites on Tezzeron and Pinchi Lakes, central British Columbia from May to October (2007-2008). Spatio-temporal variables (geographic zone, two-week time period) were used to predict the number of scats at latrine sites. 45 Observed probabilities of scat counts at latrine sites corresponded well with predicted probabilities generated from the withheld data, suggesting that the most parsimonious model had good predictive performance (Figure 6). Supporting this conclusion, a Wilcoxon ranksum test did not find a significant difference between predicted scat counts and observed counts (Z = -0.824, P = 0.41). Also, an analysis of residuals suggested very small differences between observed and predicted probabilities of scat counts. The model slightly overpredicted the number of zero counts and to a lesser degree under-predicted the number of single and double scat counts. Residuals converged towards zero as the number of scats increased. Discussion This study investigated patterns in otter diet in central British Columbia using a combination of dietary assessment techniques. Knowledge of factors influencing diet is essential for understanding the foraging ecology of otter. This understanding can help explain the mechanisms behind changes in scat deposition rates at latrine sites. Furthermore, responses to food resources will ultimately influence the distribution and density of otter populations. Management strategies require accurate and unbiased information on otter distribution and abundance that is often measured from surveys of otter sign; this study provides some of the critical information needed to achieve this objective. Results of the scat content analysis were similar to other studies in northern freshwater systems (Melquist and Hornocker 1983; Reid et al. 1994b) with some notable exceptions. Although fish dominated the diet, my study found an even smaller frequency of 46 occurrence of secondary prey items than was reported by other researchers. No single nonfish prey group exceeded 9% frequency of occurrence during any season. The insect prey group had the highest peak in frequency of occurrence in early June (8.5%), but was still much smaller than the 73% documented in northeastern Alberta in August (Reid et al. 1994b). For insects, the large difference may be explained by the techniques used to identify hard parts in scats. Reid et al. (1994b) spread hard parts out and sorted different prey groups. Using this technique, a single piece of insect remains is treated the same as fish remains that dominate the entire sample. It is also likely that insect remains are ingested secondarily while consuming fish. I used a systematic approach that incorporated a random quadrant selection process for prey identification. This method is more sensitive to the total volume of a prey item in a scat sample and, thus, does not count as many trace insect remains (i.e., lower probability of being found in a randomly selected quadrant). The same argument may be made for mollusks, but mammals and birds tended to dominate otter scats whenever they occurred in this study. Another difference between this study and past work is the prevalence of suckers in the diet of otters from Tezzeron and Pinchi Lakes. Although suckers were also dominant in the Alberta study, their peak occurrence in otter scats was similar to the average frequency of occurrence (-50%) throughout the duration of my research. Using stable-isotope analyses I was able to examine dietary separation only among distantly related taxonomic groups (with the exception of sockeye salmon). Similar isotopic values prevented an evaluation of the contribution of individual fish species to the diet of otters. The results of the first isotopic model, however, allowed the further separation of fish into more distinct groups. In general, the mixing models supported the results based on 47 visual examination of hard body parts in scats. Both approaches suggested that fish were the dominant diet item for river otter. In the first isotope model, the percentage of sockeye salmon in the diet ranged between 14-22%. Values from the second, more detailed model were consistent with this result, although the range of contribution was less broad. The results of the stable-isotope analysis suggest that marine-derived salmon food sources may be more prevalent in otter diet than was estimated in scat content analysis. There are several explanations that may account for the differences between the two analyses. Marine-derived salmon species do come into streams that are directly connected to the lakes in the study area. If otters are making shortor long-distance movements out of the study area to gain access to sockeye salmon aggregations, the scats containing sockeye salmon remains may go undetected. Alternatively, like other species, otters may concentrate on the most digestible or accessible parts of a sockeye salmon carcass when they are abundant and not consume the hard parts necessary for identification in scats (Quinn et al. 2009). This is supported by observations of otters in the study area leaving behind large fish skeletons after consuming the meat off the bone (burbot, pers.obs.). In addition, it is unclear if the scales of a spawned-out sockeye salmon pass through the digestive tract of an otter intact. When comparing results from the two analytical techniques, the largest difference involved the percentage contribution of the bird group to otter diet. In contrast to my analysis of scats where I found a peak contribution of 5%, the stable-isotope analysis suggested that this diet item may have constituted up to 27% (0-27%). Although the wide range in possible values suggests some uncertainty, the presence of birds in otter diet during 48 the summer months is supported by other studies in northern freshwater systems. Other researchers reported a peak frequency of birds in otter diet from 15 - 21.5% (Reid et al. 1994b; Melquist and Hornocker 1983; Gilbert and Nancekivell 1982). There are a couple of explanations for the differences in results between the stable-isotope and scat content analyses. First, the presence of birds in otter scats may simply occur at the lower end of the range generated by the Isosource model. Alternatively, a proportion of scats containing bird remains may go undetected when otters are eating muscle tissue and not consuming large quantities of feathers. Although the percentage of diet is uncertain, the results of this study and past work suggest that birds are an important food item for otters. Appropriate fractionation values are critical when developing stable-isotope mixing models (Tiezen et al. 1983; Milakovic 2008). I used the best available approximation available for my study population, but I was aware of the potential weakness in using a fractionation value not specific to central interior river otters. The successful use of these values in the past (Ben-David et al. 1998) and similar results from the scat content and stable-isotope analyses in this study, provide some support for the use of these fractionation values. Furthermore, a sensitivity analysis indicated that observed patterns of prey contributions were relatively robust to small changes in fractionation values. Complete validation of the mixing models I developed requires fractionation values specific to North American river otters. The timing of molt for river otters is unclear. This has implications for my results as the diet information gained from the analysis of stable isotopes is only relevant to the time period in which hair growth occurred. Guard hair is thought to regenerate from August to 49 November (Ben-David et al. 2000). For otters captured in the fall, it is unclear what stage of hair growth they were experiencing. The length of hair for otter in both spring and fall, however, was similar in length suggesting most hairs were close to fully grown. Although the scat content analysis occurred for the majority of this time period, it did not occur in November. In contrast, food sources integrated into hair during this month would have influenced the results of the stable-isotope analysis. Although some bias may have occurred, no food items that were present in otter scats went undetected in the stable-isotope analysis. In fact, the results of the stable-isotope analysis were very similar to the results of the scat content analysis with the exception of the previously noted examples. A closer examination of the ecology of prey items suggests that the minor variation in diet between techniques may be attributed to the growth cycle of otter hair. For example, the presence of the sockeye salmon and burbot/lake trout source groups in otter diet coincided with the time period for otter guard hair regeneration and the ecology of the larger fish species. Sockeye salmon and lake trout move into shallow areas to spawn during late September and October making them more vulnerable to otter predation (McPhail 2007). In addition, burbot will move into shallower areas at the mouth of sockeye salmon spawning streams in the fall (McPhail 2007). When attempting to explain variation in the occurrence of hard body parts in scats, low frequencies of occurrence for some diet items (mammal, burbot, mollusks, bird) prevented me from fitting robust logistic regression models. For other items, I had sufficient sample sizes, but the resulting models had poor predictive power (sucker, whitefish, sculpin). Although these prey items did not demonstrate predictable patterns relative to the covariates I 50 tested, important information was gained from the scat and stable-isotope analyses. Most non-fish prey items are consumed in relatively low numbers. Suckers are consumed consistently and in large quantities throughout the ice-free season, while sculpins are consumed in low numbers in all seasons. All of these prey sources are important contributing factors to otter diet, but are invariant relative to time of year and location of otter across the study area. The second ranked model for two of the analyses (presence of minnows; scat counts) had AAICC values that were <2; suggesting that they were nearly equivalent to the top-ranked model. For the count model of scat frequency, the second ranked model was the same as the top-ranked model with one additional variable. In this case, I selected the top-ranked model because the additional variable lake did not compensate for the resulting loss in parsimony (Burnham and Anderson 2002). For the model predicting the presence of minnow in otter scat, the 3 top-ranked models were nearly equivalent. I chose the highest ranked model with the fewest number of parameters (K = 7). Given the evidence of a clear best model in most cases and for ease of interpretation, I only examined coefficients for the most parsimonious model. Other spatio-temporal variables (geozone and season) from the second and third ranked models may also have influenced the number of scat at latrine sites. Models for the presence of salmonid, minnow, and insect groups had good predictive power. Kokanee probably accounted for the predictability of the salmonid group. Kokanee spawn in large numbers in streams on Pinchi Lake in late August. The location and timing agrees with coefficients that showed a significant positive influence for stream habitats, a late summer/early fall spawning period, and Pinchi Lake. Kokanee are seasonally abundant and 51 high energy food sources. A radio-telemetry study in Idaho documented long-distance movements that coincided with kokanee spawning seasons (Melquist and Hornocker 1983). In addition, a concurrent radio-telemetry study of otters in this study area documented extended movements by two river otters from Tezzeron Lake to streams on Pinchi Lake when kokanee were spawning (pers. obs.). These long-range movements suggest that seasonally abundant prey sources are important having significant influences on the distribution and abundance of otters across the landscape. As with other species, habitats where kokanee spawn may be particularly important to otter ecology. The sound management of healthy kokanee populations and their habitat will have direct benefits to river otter populations. The model for the minnow prey type also had good predictive accuracy. Model coefficients suggested that otter in Tezzeron Lake feed more consistently on minnows. The differences in diet between the two lakes highlight the need to develop predictive models that incorporate appropriately large geographic variation. The spawning period also had a positive influence on minnow occurrence in scat samples, with a greater prevalence during the late September - October time period. This time period does not correspond with the spawning season of the minnow group and is not easily explained by minnow ecology. Reid et al. (1994b), however, found that the occurrence of minnows increased in the fall and dominated the diet during the winter months. For insects, the positive influence of the early June and late June - early August time periods is not surprising given that many insects emerge in large numbers during this time. Although insects may be overestimated in otter 52 diet, large numbers of scats containing only insect remains suggest that this prey source is not purely incidental. I developed a count model to investigate the relationship between otter behaviour at latrine sites, as indexed using scat numbers, and variables that captured the spatio-temporal variation in the availability of prey groups. The top-ranked ZINB count model included covariates for geographic zone and time measured in two-week periods. Although numbers of scats at a latrine sites were influenced by different geographic locations across the lakes, this variation could not be explained ecologically. Assuming that the relative abundance of scats reflects relative use by otters, this result suggests that otters use different areas on the lake with differing intensities. The effects of prey distributions on otter activity and movements will strongly impact the results of surveys designed to measure and manage otter populations. Survey protocols for input into management plans need to take into account the spatio-temporal variability associated with otter activity. The positive influence of the early July time period on scat numbers did not correspond with the spawning periods of any fish groups. Furthermore, the spawning period of fish, with the exception of kokanee, did not have a significant affect on the number of scat deposited at latrine sites. One possible explanation for the seasonal trend is that prey species diversity and abundance increases in the summer (Reid et al. 1994b), leading to increased movement, social grouping, and use of latrine sites. During the summer months otters will often travel in family or bachelor groups while foraging (Reid et al. 1994b; Gorman et al. 2006). A concurrent radio-telemetry study on Tezzeron and Pinchi Lakes also found that marked otters traveled together during the summer, and observations of groups with 3 to 7 similar-size otters were not uncommon (pers. 53 obs.). Groups of this size suggest the presence of an abundant food supply may explain the increase in scat deposition at latrine sites in mid-summer. The early September increase in otter activity at latrine sites was most likely explained by concentrations of spawning kokanee. This explanation is consistent with an observed increase in the occurrence of the salmonid family during this time period. Once again, these results have implications for otter surveys and monitoring strategies. For example, the power of mark-recapture studies increases when the chance of a successful detection is maximized (Lukacs and Burnham 2005). This study investigated river otter ecology and diet in central British Columbia using a combination of techniques. The paired application of scat and stable-isotope analysis provided an opportunity to compare results and offset biases associated with each approach. Corresponding results suggest that in this case the two techniques have similar utility for describing the diet of otter. From a sampling efficiency perspective, the stable-isotope approach offers considerable time savings, contingent on a ready supply of otter hair or other tissue. The successful use of stable isotopes in this study and publication of baseline C-N values will be useful for the application of this approach to other populations of river otter. Although the results of this study are specific to the Tezzeron and Pinchi watersheds, they should have applicability to other areas of central British Columbia that have similar habitats and prey resources. Before this study, the predictability of otter diet and scat deposition at latrine sites had not been investigated. Predictive models of otter diet could be especially useful in studies that span all seasons and/or occur in areas with a temporally more variable and diverse prey 54 base. Prey distribution and availability may be the strongest influences on river otter movements, distribution, and relative abundance. Surveys of latrine site locations are not easily interpretable without knowledge of the mechanisms behind variations in their use. The ability to predict otter diet and latrine use can be used by natural resource managers to mitigate the effects of harvest or land development on otter populations. An understanding of spatio-temporal patterns of diet and latrine site activity is essential for interpreting surveys that use sign to assess the distribution or abundance of otter populations, and for developing management strategies that maintain the prey base for otter. important component in addressing this need. 55 This study provides an CHAPTER 3 Spatial and behavioural scales of habitat selection and activity by river otters at latrine sites. Abstract Animals interact with their environment at multiple spatial, temporal, and behavioural scales. Few studies of selection for latrine sites by river otters (Lontra canadensis) have considered spatial scale, and no studies have integrated scales of behaviour. I used an Information Theoretic Model Comparison (ITMC) approach to identify elements of otter habitat that influence the presence, consistency, and intensity of latrine site activity at two spatial scales. I identified 73 latrine sites on Tezzeron and Pinchi lakes and their associated tributaries during intensive shoreline surveys in 2007 and 2008. Latrine sites were surveyed every two weeks for two years during the ice-free season to monitor visitation rates. I inventoried latrines and randomly selected sites along the adjacent shoreline, and used those data in the form of a binary Resource Selection Function (RSF) to model fine-scale selection of latrine sites. At the scale of the landscape, I used an RSF and data from Geographic Information Systems (GIS) to model coarse-scale selection of latrine sites. Drawing on those same data, I used binary and count models to quantify factors that contributed to the consistency (high vs. low use) and intensity (number of scats) of otter activity at latrine sites. Fine-scale habitat characteristics were better at predicting the presence of latrine sites when compared to coarse-scale GIS data. In general, the presence, consistency, and intensity of latrine activity at the fine-scale were influenced by visual obscurity, larger trees, and characteristics of conifer trees. The presence of latrine sites at the coarse scale could not be accurately described by any of the models. The consistency and intensity of activity of otters at latrine sites at the coarse scale, however, was best predicted by habitat characteristics beneficial to fish. These results provide insight into the spatial and behavioural scales of 57 latrine site activity by river otters that can be incorporated into management, monitoring, and conservation strategies. Management decisions are often based on estimates of population trends, and this study provides the baseline information required for developing techniques for assessing and monitoring otter populations and their habitat. Introduction The inclusion of temporal and spatial scales in ecological studies is critical to the interpretation of resource selection (Johnson 1980; Wiens 1989). Variables at the coarse scale may be missed or fine-scale patterns may be averaged depending on the nature of the measurement (Dunning et al. 1992). In addition, by using scale to delimit behaviour we can begin to infer mechanisms that drive resource selection (Johnson et al. 2002). A growing number of studies have begun to investigate behavioural and spatial scale using detailed movement data and Global Positioning Technology (GPS) (Johnson et al. 2002; Fritz et al. 2003; Frair et al. 2005). Variation in the amount of sign (i.e., scats, tracks) at sites used by animals also has the potential to reveal spatial and behavioural scales of habitat selection. For example, North American river otters (Lontra canadensis) visit latrines and leave behind scat that is both identifiable and measurable through space and time. From a scalar perspective, the selection and use of latrine sites by otters is likely a trade-off between accessibility of prey resources at the coarse scale and access to adequate cover at the fine scale. Latrine sites are areas where otters consistently come on to shore to deposit feces, scent-mark, and roll around in terrestrial vegetation and debris. Ben-David et al. (2005) 58 investigated a coastal population of river otter and found different functions for latrine sites depending on the gender and social status of individuals. Social otters most likely scent marked for intra-group communication, non-social otters to signal mutual avoidance, and females for defense of territories. Rostain et al. (2004), using captive otters, suggested that feces are deposited at latrine sites to communicate social status. In Eurasian otters (Lutra lutra), latrine sites may signal the active use of food resources (Kruuk 1992). Regardless of their function, latrine sites are used widely by otters, serve an important role in otter ecology, and are easily identifiable along shorelines. The primary use of latrine locations by researchers has been to determine occupancy, distribution, and habitat selection (Dubuc et al. 1990; Newman and Griffin 1994; Swimley et al. 1998). Tributaries, points of land, coniferous trees, rock formations, and fallen logs commonly characterize latrine sites (Dubuc et al. 1990; Newman and Griffin 1994; Swimley et al. 1998). Beaver activity has also been reported as an important factor describing otter habitat use (Melquist and Hornocker 1983; Dubuc et al. 1990; Swimley et al. 1998). These studies assume latrine sites are an accurate predictor of river otter habitat use and distribution. There is very little information, however, relating the location and habitat features of latrine sites to spatial and temporal variation in their use. The exceptions are the use of deposition rates at latrine sites to document differences in latrine use between lake and stream habitats during the spring and summer (Crait and Ben-David 2006), and to investigate coarse-scale differences in the use of beaver ponds or wetland types (Leblanc et al. 2007; Newman and Griffin 1994). Variation in latrine-site use among seasons has also been documented for Eurasian otters in Finland (Sulkava 2006). 59 Failure to investigate spatio-temporal patterns in visitation rates can have implications for monitoring protocols and limits our understanding of otter ecology. Intraseasonal variation in the use of latrine sites will influence the success of surveys designed to document the location of latrines. For example, surveys conducted in spring or early summer may not detect latrine sites that are still under water. Furthermore, latrine surveys conducted during a time of year when an influential prey source is abundant may bias the observation of active latrines. For example, latrines far from salmonid spawning areas my receive little activity during this time of year and go undetected during surveys. Latrine sites found during short-term surveys often serve as indicators of habitats used by otters (Dubuc et al. 1990; Newman and Griffin 1994; Swimley et al. 1998). Temporal differences in the use of latrine sites by otters, however, could dramatically affect interpretations of habitat selection. In addition, there can be extreme variations in the number of scats found at latrine sites. Typically, the only criterion for documenting a latrine site location is the presence of a minimum number of scats. For example, when documenting habitat selection of river otters in Pennsylvania, Swimley et al. (1998) required that a latrine site have > 2 scats. Using this criterion, a latrine site that has only 2 scats and was visited only once during the season is treated the same as a latrine site used throughout the entire season with dozens of scat deposits. Some latrines may have a long history of frequent visitation, while others may be just an ephemeral site used by an otter at a single point in time. Such variation in use is likely the product of important ecological determinants of otter distribution and perhaps abundance; however, there is very little information relating habitat characteristics at latrine sites to the consistency and intensity of activity by otters. 60 Information relating latrine habitat characteristics to the degree of otter activity would be beneficial to natural resource professionals prioritizing management decisions that emphasise the habitat requirements of otters. Incorporation of scale into ecological studies is essential and may have profound affects on observed patterns of species distribution (Wiens 1989). Studies of latrine sites in the past have typically focused on either fine- or coarse-scale habitat characteristics or have combined characteristics across scale (Dubuc et al. 1990; Bowyer et al. 1995; Swimley et al. 1998). Selection by otters for latrine sites, however, may be driven by very different environmental factors depending on the scale of the observation. Frequently, habitat selection by wildlife species is strongly correlated with the presence and distribution of food resources. At the coarse scale (landscape), I hypothesize that otters will select latrine sites based on habitat characteristics that affect prey distributions. At the fine scale (shoreline patch), I hypothesize that otters will select areas with characteristics that provide vertical and horizontal security or environmental cover. There are two reasons why cover may be important for otters. First, otters may be vulnerable to predation by avian or terrestrial predators when transitioning from the water onto land. Although only anecdotal information exists, gray wolves (Canis lupus), black bears (Ursus americanus), brown bears (Ursus arctos), and bald eagles {Haliaeetus leucocephalus) are potential predators of otters that occur in the study area (Melquist et al. 2003). Second, the cover may protect scent from environmental influences, such as rain and sun, prolonging the impact of scent-marking behaviour. 61 In this study, I investigated selection by river otters for latrine sites at multiple spatial and behavioural scales. I used an Information Theoretic Model Comparison (ITMC) approach to identify elements of otter habitat that influenced the presence, consistency, and intensity of latrine site activity. I identified and inventoried latrine sites, and adjacent control locations, on Tezzeron and Pinchi lakes and their associated tributaries in central British Columbia. Latrine sites were surveyed every two weeks for two years during the ice-free season to monitor visitation rates. I used the data describing the latrine and matched random control sites to develop fine-scale binary Resource Selection Functions (RSF). At the scale of the landscape, I used Resource Selection Functions and data from Geographic Information Systems (GIS) to model coarse-scale selection of latrine sites. Drawing on the visitation data, I used binary and count models to quantify factors that contributed to the consistency (high vs. low use) and intensity (number of scats at latrine sites) of otter activity at latrine sites. The objectives of this study were to: (1) identify habitat characteristics that influenced the selection of latrine sites at two spatial scales; and (2) investigate the activity patterns of otters relative to the habitat characteristics of latrine sites. Methods Data Collection I identified latrine site locations through a series of shoreline surveys on both Pinchi and Tezzeron Lakes as well as along all significant tributary streams (1 km from lake-stream confluence and navigable by canoe or kayak). I conducted shoreline surveys by canoe, kayak, and on foot. Two complete surveys of all shorelines were conducted in 2007. The 62 first survey occurred from 5-27 June and the second from 20 July to 5 August. From 15 August to 15 September, 2008, I randomly selected and intensively surveyed 200, 200-m segments of shoreline split evenly between the Tezzeron and Pinchi lake systems. The 2008 survey was conducted to both identify new latrine sites and to determine if the majority of the active latrine sites were being monitored. I chose to conduct surveys during three distinct time periods, late spring, summer, and early fall, to account for differing prey availability and biological constraints. In late spring, suckers and trout species move into stream systems to spawn and the movements of female otter are restricted by the presence of offspring. In summer, prey diversity is at its highest and adult female movements are less restricted as pups become more mobile and leave the natal den. In fall, kokanee (Oncorhynchus nerka) and sockeye salmon (Oncorhynchus nerka) move into streams, pup mobility is at its highest, and female movements are least restricted. After latrine sites were identified, I surveyed each site every two weeks to collect scats and record the number of scats deposited. In 2007, latrine site monitoring began in July and ended in late October. In 2008, latrine site monitoring began in mid-May and ended in mid-October. Fine-scale Selection and Activity A latrine site was included in the habitat selection analysis only if it contained >3 scats combined across all visits, and was visited >3 times during the duration of the study. I used a 1:1 sampling design to generate a paired random site as a control for each latrine site. I used a random number table to locate the random site between 21-100 m along the adjacent 63 shoreline from the reference latrine site. Plots at latrine and non-latrine sites consisted of a 5.64-m diameter half circle. Plots at latrine sites were centered on the most frequently used entrance trail from the water. All plot centers were located 1 m inland from the origin of terrestrial vegetation and perpendicular to the shoreline. Habitat measurements included: visual obscurity, percent cover, conifer cover, bank height, slope, substrate, and tree characteristics (Appendix 2). Data collection protocols were guided by the BC Vegetation Resources Inventory Guidelines (Resources Inventory Committee 2006). Visual obscurity was measured using a cover pole at 5 m inland from the shoreline. Percent cover for all vegetation layers was estimated to the nearest 5% using an ocular estimate. Tree distances were measured from the tree trunk to the edge of the terrestrial vegetation line. Tree dripline was measured from the tree trunk to the outer edge of the longest branches in the direction of the water. Diameter at breast height (dbh) was measured for all trees >7.5 cm. Percent slope was measured within the latrine site using a clinometer (Suunto PM-5). Tree dbh, distance, and drip-line extent measurements were grouped into categories based on the distribution of measurements, plot size, and potential ecological significance of cover attributes (dbh = 0-29, 30-49, >50 cm; tree distance = 0-1.9, 2-3.9, >4 m; dripline = 0-.9, 11.9, 2-2.9, >3m). Coarse-scale Selection and Activity For coarse-scale selection of latrine sites, random points (n = 200) and associated spatial data were generated using ArcGIS (ArcMap ver.9.3 by ESRI). Variables for coarsescale selection represented aquatic habitat relative to fish ecology and characteristics of 64 terrestrial vegetation. Coarse-scale variables included: distance to beaver lodge, distance to reed patch edge, distance to navigable fish-bearing stream mouth, and water depth (100 m from shoreline). Areas with beaver ponding are often rich sources of biomass because of the productive invertebrate and fish habitat created by beaver structures (Gard 1961; McDowell and Naiman 1986). Reed patches create structure that may be important rearing, foraging, and escape cover for many fish species. Stream mouths are often productive areas, especially during times when a fish species moves from deeper lake water into shallower streams to spawn. Otter cannot productively forage in water that is very deep and water depth may affect the vulnerability of different prey species (Melquist et al. 2003). Water depth was measured directly from 1975 bathymetry maps created by the BC Ministry of the Environment. Vegetation measurements were taken from provincial Vegetation Resources Inventory data (Resources Inventory Committee 2008) and included: dominant tree species, percent dominant tree species, average tree height, and canopy closure. I included vegetation characteristics that describe cover because they may also have an influence on the locations of latrine sites at the landscape level. Variables such as canopy cover or dominant species may provide vertical or horizontal cover for otters allowing them to avoid predators or prolong scent influencing the location of latrine sites at the landscape scale. Habitat Models I developed three types of predictive models to investigate spatial and behavioural scales of resource selection of latrines and activity by otters. The first model, latrine site selection, investigated habitat characteristics that influenced the presence of latrine sites. The 65 other two models representing the consistency and intensity of use, investigated the importance of habitat characteristics on the visitation behaviour of otters at latrine sites. Consistency of use was measured as the number of times that a latrine site was observed as active. I considered a site active if an otter had deposited scats, regardless of the level of disturbance or number of feces at the latrine site. Intensity was measured by counting the number of scats deposited at latrines sites during each survey. Consistency measured how often a latrine was active, while intensity measured the amount of activity when it was active. All three model types were conducted at two spatial scales, the landscape and shoreline patch. Selection of Latrine Sites I used a binary Resource Selection Function (RSF) to investigate habitat characteristics that influenced the selection of latrine sites by otters. I used a presenceabsence design and conditional fixed-effects regression to develop a set of models to investigate selection at the scale of the shoreline patch. In contrast to conventional logistic regression, the conditional fixed-effects model takes into account matched groups (Hosmer and Lemeshow 2000). At the scale of the landscape, I used a Resource Selection Function (RSF) and data from Geographic Information Systems (GIS) to model coarse-scale selection of latrine sites. For this analysis, the entire shoreline was considered available, thus, I used a conventional logistic regression, not a matched design. 66 Latrine Consistency The influence of habitat characteristics on the consistency, or the number of visits by otter to a larine site, was modeled using conventional logistic regression. Latrine sites were split into low/high use categories based on the number of surveys in which a latrine site was observed as active. A latrine site was considered to be high-use if it was active >65% of the times it was surveyed. This delineation was based on the distribution and median value of the data set. Observations of monitored latrine sites indicated that a >65% value was a conservative delineation for defining a high-use latrine. Latrine Intensity I used the number of scats counted at a latrine site during a survey event as an index of the intensity of use by otters. I used these count data and a zero-inflated negative binomial (ZINB) model to investigate the influence of habitat characteristics on the intensity of activity at each latrine site. A ZINB accounts for both over dispersion and the presence of excess zeros in the data set. I used a likelihood ratio test to confirm that the Negative Binomial distribution was preferable to the Poisson. I then used a Vuong test to determine if a zero-inflated model was required (Vuong 1989). All data analyses were performed using Stata (ver. 9.2, Statacorp, 2006). Habitat Model Development Eighteen different variables were used to develop models for the fine-scale selection of latrine sites (Table 9). Variables were a combination of vegetation characteristics and 67 shoreline topography. I used eight variables in the development of models to explain coarsescale selection of latrine sites (Table 10). None of the parameters for the fine-scale analysis were expected to have non-linear distributions, so quadratic equations were not required. Many of the distance values in the coarse-scale selection model, however, would be expected to display a non-linear distribution. Parameters such as distance to beaver lodges, reed patches, and stream mouths were tested with and without a quadratic term. I used deviation coding (desmat.ado; Hendrickx 2001), in which the effect of each variable is compared with the overall mean effect of the independent variable, to represent the categorical variables (Menard 2001). I developed a total of nine models as hypotheses to explain the presence or absence of latrine sites (Table 11), and six models to explain the consistency and intensity of use of individual latrine sites at the scale of the shoreline patch (Table 11). Models were combinations of vertical or horizontal cover, all tree characteristics, conifer characteristics, and/or shoreline topography that may be important to otters because of their vulnerability to predators or scent-marking behaviours. Habitat variables such as visual obscurity, percent cover, or conifer trees may provide cover that reduces exposure to terrestrial predators or protects scent-marking areas from environmental influences. Topographical features such as bank height and slope may increase access to areas that remain available for scent-marking throughout the ice-free season (i.e., during spring flooding). For the coarse-scale selection of latrine sites, I developed a set of eight biologically plausible models for the binary and count analyses (Table 11). 68 Table 9. Parameters used in the development of binary and ZINB count models for the selection of latrine sites and activity by river otter, based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia, from 2007-2008. Parameter herb shrub1 shrub2 tree shltree distance dbh obscuritytot obscurityll5 slope bankheight substrate spruce subfir birch willow conifer dripline condistance condbh Description % herb (<15cm) % shrub (2-10m) % shrub (0-2m) % tree (10+m) % shrub 1 and tree combined average distance of trees to vegetation line (m) dbh max of all trees (cm) % visual obscurity total (0-1.5m) % visual obscurity (l-1.5m) % slope within latrine site bankheight (cm) substrate between water and vegetation line # spruce trees # subalpine fir # birch trees # willow trees # conifer trees average conifer drip line distance (m) average conifer distance to water (m) maximum conifer dbh (cm) 69 Variable type continuous continuous continuous continuous continuous categorical categorical continuous continuous continuous continuous categorical continuous continuous continuous continuous continuous categorical categorical categorical Table 10. Parameters used in the development of binary and ZINB count models for the selection of latrine sites and activity by river otter, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia, from 2007-2008. Parameter tree height dspecies canopycover dspecies % dbeaverlodge dbeaverlodge2 dreedpatch dreedpatch2 dstreammouth dstreammouth2 waterdepth Description tree height dominant tree species canopy cover % dominant tree species distance to nearest beaver lodge distance to nearest beaver lodge squared (quadratic) distance to nearest reed patch edge distance to nearest reed patch edge squared (quadratic) distance to nearest fish-bearing steam mouth distance to nearest fish-bearing stream mouth squared (quadratic) water depth measured 100m perpendicular to shoreline 70 Variable type continuous categorical continuous continuous continuous continuous continuous continuous continuous continuous continuous a o ^ t n v o i n « > n ^ > h n in m o o\ h oo oo £ <* T3 CS o 00 1/1 & '-2 03 'H-> 43 oo !-T 4) 'to e 4> ^ >-> o a o H—> 00 4> > o o O > o o ^3 i o 03 VH VH .3 3 4> 43 00 U-l O o 4-> 03 03 o 43 1 43>-. a CJ 4= OH o o3 03 03 4 3 o '-H f, U I-! 43 O o 60 O c •c—( '-S £ H U a •d o 03 4) 60 5 > u _ >O H K M U ro o n oo g X) 'S 44>3 «* o '£ 4) > o o D "53 O o o o *-• O H _g 3 « ~0 ' C 4> > 4) -c 00 en 1) o (D u a33 -t-» oo 4) 60 4) 3 > abi >, -4—' o 43 XI •a >-> X) oo o o 03 53 o 1-H u 4) 43 _S3 13 l-H o O ^3 C/3 < 60 4) 60 4) 4) 3 4) C > > quati lobal > 4) <4-l fer fer -j >- o o 43 43 00 1/5 O o \-i 00 e o <« Q ^p seni o of •— & >> -4-4 |> '& T3 3 03 oo flj '5 5 IS a "o U X! oo 'C 43 O m H—» 13 is OH 4) O 4) O H > U, 03 T3« 3 -51 W >. 3 J 43 •d • 1—1 -a 4> ea u fi <+H election o OO ort 43 <+H 43 < 'T3£ 3 03 3 O I-I N N) 4) H 3 O 00 T3 13 73 1) o a 4> -4-» OS T3 ^ 03 a riori o ft "3 4) "o 03 03 -a u •a 00 1 4> 00 S3 4> oo 4> ? > 43 o 03 _,' 'S kH o + a 1 43 00 • ^ H )-H o 60 u + kH X> ,"H-H U 2 'S 43 o u 03 £ t5 303 i H e o a 03 o3 OH a o o •d 3 + in GO O O + + o c u a 03 T3 T3 43 + + US c '53 43 en + + + s .4) 4) 43 43 o + • ~ H 43 + C 43 4 3 M 4) O 3 m o + i d —i " ^ CN 43 3 ^ + -2 ± ^ + 'C 4) O T3 2 OH „4) ?! 5 1/3 00 T3 + o3 43 60 4) 03 -g + 60 4) J * J3 -o 00 g o3 ^ 1 a •£ 1 3 S oo ^ -O ^ nb ± «^ T 4) 00 4 3 ^ T S S S ^ o\J2 > -3 > + 4) VO + O - f 73 4) P , M ft & a S oo 43 O + • 5 43 +I 03 ,4> -a 2 03 4) U C 00 + 4) * 2 OH + OH 2 T3 6 0 ^ T 3 O O 4) 0-43 a >^ g O 43 o 3 4) - S OH "o3 OH ^ 00 o< 42 O n i 2 4) T? 4) 5S C 03 43 + a + x « + 2 c a S *- e •? + UH CO 43 43 •a T3 3 o 4) 5 00 £;€ + 3 2 43 O + CA) 43 oo 43 + 03 ao rt IT) U (U w< kH + + ,U "S o j* "^ 43 4) 4) 43 s 00 >-H a 43 4> o O T3 C y <^ 5 + l-H t-H ^ H ^ H « 40 O o^ 03 u £ ° u >> >^ 43 £ 43 T > o"^ X QH OH O 53 O 53 43 a + + .SPlS .2? "2 S O S > «, o o u 43 03 4 3 on 4) a) H 8 ft -O ^ &1> 4 3 4> X nb 73 nb £ 00 00 T3 T3 Global, terrestrial vegetation, and aquatic habitat models were developed as well as two models that combined terrestrial vegetation variables with different aquatic habitat attributes. Terrestrial vegetation may be important for cover and aquatic habitat may influence the prey distributions and abundance of prey sources. I used variance inflation factors (VIF) to assess each variable for excessive multicollinearity. I removed variables from a model if they had a VIF value greater than 10 or a mean VIF value greater than 1 (Chatterjee et al. 2000). In this study, none of the model variables used in the analyses had excessive multicollinearity (i.e., VIF>10). Habitat Model Selection I used Akaike's Information Criterion (AICC) for small sample sizes to identify the most parsimonious explanatory models of latrine selection and activity by otters (Burnham and Anderson 2004). The AICC values are a relative metric that must be compared in the context of a set of a priori models. I used both AAICC and Akaike weights (AICw) to rank and compare models. The model with the lowest AICC score is considered the "best" or the most parsimonious model given the data and the set of models compared. A model with a AAICC <2, however, was considered to be equivalent to the model with the minimum score (Burnham and Anderson 2002). When models had AAICC values that were nearly equivalent, I selected the most parsimonious model (i.e., fewest number of parameters). An AICw is a value from 0-1 that represents the approximate probability that a model is the best among a set of candidate models. I used beta-coefficients and z-statistics (P < 0.05) to assess the 72 importance of individual parameters contained within the most parsimonious explanatory models. Predictive Ability of Habitat Models Data for all count models were randomly divided into training (85%) and testing (15%) groups using a random number generator and a uniform distribution. The count models were developed using the data training group and then validated using the data testing group (Fielding and Bell 1997). I located too few latrine sites to conduct an independent evaluation of the predictive accuracy of binary models. I used the receiver operating characteristics (ROC) and resulting area under the curve (AUC) to assess the predictive ability of the "best" model from the binary analyses. The AUC measures the relative proportions of correctly and incorrectly classified prediction (Pearce and Ferrier 2000). AUC values 0.5 to 0.7 were considered to have poor model accuracy, from 0.7 to 0.9 good model accuracy, and >0.9 were considered to have high model accuracy (Swets 1988). I used Pearson's standardized residuals to identify outliers (Menard 2001). I used the predicted counts as well as the predicted probabilities of counts to evaluate the predictive performance of the most parsimonious count models (prcounts.ado; Long and Freese 2006). I evaluated the performance of the model by visual inspection of graphs plotting the observed probability of a count using the model testing data and the predicted probability of a count generated from model training data. The residual difference between observed and predicted counts allowed me to further examine the models predictive ability across the range of values I observed. 73 Results I located a total of 73 latrine sites across 155 km of shoreline (Figure 1). Sixty-seven and six latrine sites were found in 2007 and 2008, respectively. Only two new latrine sites were found in areas already surveyed in 2007. The other four latrine sites were found in areas of Tezzeron Creek not surveyed in 2007. The 2008 survey was conducted to validate latrine locations found in 2007. The results demonstrate that I was monitoring the majority of latrine sites found along shorelines in the study area. Because of time constraints, I was able to conduct habitat measurements at 70 latrine sites only. Selection of Latrine Sites The tree characteristic/cover model best explained the presence of latrine sites at the scale of the shoreline patch (Table 12). The second and third ranked model had some support but had AAICC scores that were 3.5 and 3.6 points, respectively, greater than the firstranked model and were not considered equivalent. Vertical cover >2m and the number of conifer trees were two variables that were common to the three top models. The ROC score showed that the top-ranked model had good predictive accuracy (AUC = 0.821). A maximum tree dbh > 50 cm had a positive statistically significant influence on the presence of latrine sites, while a tree dbh < 29 cm had a negative influence. Visual obscurity (1-1.5 m) had a significant positive influence on the presence of latrine sites (Table 13). The shoreline vegetation/spawning habitat model best explained the presence of latrine sites at the coarse scale. The Akaike's weight indicated that the top-ranked model had 74 a 74% chance of being the best among the candidate models. The model, however, had poor predictive accuracy (AUC = 0.658). Latrine Consistency The conifer characteristics/horizontal cover model best explained the consistency of otter activity at latrine sites (Table 12). The Akaike's weight indicated that the top-ranked model had an 89% chance of being the best among the candidate models; this model also had good predictive accuracy (AUC = 0.765). The number of spruce trees, visual obscurity (11.5 m), and extent of conifer drip line had a positive influence on the consistency of latrine site activity (Table 13). At the coarse scale, the aquatic habitat model best explained the consistency of activity at latrine sites (Table 14). The aquatic habitat model had good predictive accuracy (AUC = 0.744). The distance to beaver lodge had a statistically significant negative influence on the consistency of use of latrines by otter, while the distance to reed patch had a positive influence (Table 15). Latrine Intensity For the analysis of intensity of use of latrines at the fine scale, a negative binomial regression (NBRM) model performed better than a Poisson model (PRM) (G2 = 2007.28, P < 0.001), and because of the large number of zeros in the data set a ZINB provided a better fit than a NBRM (Vuong = 2.91, P < 0.002). 75 Table 12. Summary of AICC model selection statistics for candidate models (binary and ZINB count) predicting latrine selection and activity (occurrence, consistency, and intensity), based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008. Model Name Latrine Selection Binary Model Cover/tree characteristics Global Vertical cover Cover and tree distance Conifer cover/tree characteristic Tree species Horizontal cover Tree characteristics Topography Consistency Binary Model Conifer characteristics/horizontal cover Topography Vertical cover Conifer characteristics Tree characteristics/cover Tree characteristics Intensity ZINB Count Model Tree characteristics/cover Conifer characteristics/horizontal cover Conifer characteristics Vertical cover Topography Tree characteristics Rank AICC AIC C A AICw 1 2 3 4 5 6 7 8 9 57.5 61.0 61.1 66.0 69.8 72.6 73.7 79.5 89.0 0.0 3.5 3.6 8.5 12.3 15.1 16.2 22.0 31.5 0.738 0.127 0.122 0.011 0.002 <0.001 <0.001 <0.001 <0.001 1 95.5 0.0 0.887 2 3 4 5 6 101.8 101.8 102.2 106.8 107.7 6.3 6.3 6.7 11.3 12.2 0.039 0.037 0.032 0.003 0.002 1 4352.6 0.0 0.923 2 3 4 5 6 4357.9 4361.2 4394.8 4399.8 4402.0 5.3 8.6 42.2 47.2 49.4 0.065 0.012 <0.001 <0.001 <0.001 76 Table 13. Estimated coefficients for AICC selected models (binary) predicting the selection of latrine sites and consistency of activity by river otters, based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008. Parameter Coef. SE 95% CI Latrine selection (cover/tree characteristics model) dbh (0-29 cm) 0.546 -1.128 -2.198 —- -0.058 dbh (30-49 cm) 0.442 -0.353 -1.219 —-0.513 0.630 dbh (50 cm+) 1.481 0.246 --2.716 shrub 2 0.018 -0.064 -- 0.006 -0.029 conifer 0.144 0.219 -0.063 --0.501 0.012 shrub 1 tree 0.023 -0.001 -- 0.047 0.028 0.012 0.004 —- 0.052 obscurity (1-1.5 m) Consistency (conifer characteristics/horizontal cover) spruce 0.185 0.430 0.067 -- 0.793 0.608 dripline (0-.9 m) 0.793 -0.399 --1.985 dripline (1-1.9 m) 0.576 -0.897 -2.026 -- 0.232 0.566 dripline (2-2.9 m) -1.053 - 2 . 1 6 2 - - 0.056 dripline (3 m+) 1.157 0.578 0.024 —- 2.290 shrub2 0.016 -0.017 -0.048 --0.014 obscurity (1-1.5 m) 0.018 0.009 0.000 -- 0.036 constant -1.015 0.930 -2.838 --0.808 77 P 0.039 0.424 0.019 0.112 0.129 0.053 0.021 0.020 0.192 0.120 0.063 0.045 0.300 0.049 0.275 Table 14. Summary of AICC model selection statistics for candidate models predicting latrine activity (consistency and intensity) by river otters, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes in central British Columbia from 2007-2008. Model Name Consistency Binary Model Aquatic habitat Shoreline vegetation/fish and beaver habitat Shoreline vegetation characteristics Shoreline vegetation/ spawning habitat Global Intensity ZINB Count Model Global Shoreline vegetation characteristics Shoreline vegetation/spawning habitat Aquatic habitat 1 Shoreline vegetation/ fish and beaver habitat Rank AICC AICcA AlCw 1 2 3 4 5 97.6 99.9 104.0 104.3 107.4 0.0 2.3 6.4 6.7 9.8 0.788 0.311 0.040 0.035 0.001 1 2 3 4 5 4143.5 4173.2 4175.1 4178.7 4188.5 0.0 29.7 31.6 35.2 45.0 0.999 <0.001 <0.001 <0.001 <0.001 Table 15. Estimated coefficients for the AICC selected model (binary) predicting the consistency of activity by river otters at latrine sites, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008. Parameter Coef. Aquatic Habitat Model dbeaverlodge 2.357 dbeaverlodge -0.001 dreedpatch -1.812 2 dreedpatch <0.001 dsstreammouth -0.684 dstreammouth <0.001 waterdepth 0.053 constant 0.432 SE 95% CI P 1.057 <0.001 0.802 <0.001 0.603 <0.001 0.040 1.044 0.285 — 4.429 -0.001—-0.001 -3.384 —-0.240 <-0.001— <0.001 -1.866 — 0.498 <-0.001—<0.001 -0.025 — 0.131 -1.614 — 2.478 0.023 0.054 0.045 0.087 0.109 0.176 0.062 0.901 78 Table 16. Estimated coefficients for the AICC selected model (ZINB count) predicting the intensity of activity by river otters at latrine sites, based on fine-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008. Parameter SE 95% CI P 0.079 -0.030 -0.050 0.137 -0.218 0.081 0.037 -0.005 -0.014 -0.003 2.468 0.123 0.089 0.093 0.111 0.115 0.160 0.028 0.002 0.005 0.003 0.350 -0.161- - 0.320 -0.205 --0.145 -0.232 --0.133 -0.803 --0.354 -0.443 -- -0.008 -0.232 -- 0.394 -0.017- -0.091 -0.010 —- -0.001 -0.023 —- -0.005 -0.009 -- 0.004 1.782- -3.315 0.517 0.739 0.594 0.217 0.048 0.613 0.181 0.037 0.002 0.403 <0.001 10.282 -1.137 -9.145 10.830 9.947 -20.777 2.488 -0.300 -0.112 -0.308 3.636 1.419 0.298 1.243 1.020 0.948 1.928 0.408 0.041 0.017 0.046 1.043 -7.500 — 13.065 -1.721 —--0.553 -11.582—- -6.709 -8.832— -12.828 8.089 — 11.805 -24.556 -- 16.997 1.689 —-3.287 - . 0 3 8 1 - - -.0220 -0.145 —- -.0780 -0.398 ---0.219 1.592- - 5.680 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 Coef. Tree characteristics/cover model dbh (0-29cm) dbh (30-49cm) dbh (50cm+) distance (0-1.9m) distance (2-3.9m) distance (4m+) conifer obscurity (1-1.5m) shrub2 shltree constant Inflate Portion dbh (0-29cm) dbh (30-49cm) dbh (50cm+) distance (0-1.9m) distance (2-3.9m) distance (4m+) conifer obscurity (1-1.5m) shrub2 shltree constant 79 I found a similar result for the coarse-scale analysis; an NBRM model performed better than a PRM model (G2 = 1893.34, P < 0.001),and a ZINB provided a better fit than an NBRM (Vuong = 3.56, P< 0.001). The tree characteristics/cover model best explained the intensity of latrine site activity by river otters at the fine scale (Table 12). The average tree distance (2-3.9 m), the visual obscurity (1-1.5 m), and the shrub cover (0-2 m) had a negative influence on the number of scats at latrine sites (Table 16). At the coarse scale, the global model best explained the intensity of latrine site activity by river otters (Table 14). The Akaike's weight indicated that the top-ranked model had a near 100% chance of being the best among the candidate models. Latrines with a large amount of otter activity were associated with stream mouths and reed patches as shown by negative and significant coefficients for both distance variables (Table 17). At the fine scale, the count model describing the intensity of use of latrine sites resulted in a good fit between the observed and the predicted probability of the number of scats (Figure 7). Furthermore, the Wilcoxon rank-sum test did not find a statistically significant difference between the predicted and observed counts (Z = 1.126, P = 0.26). The mean of the residual analysis was close to zero (X < 0.001), with residuals converging towards zero as the number of scats increased. The count model describing latrine intensity at the coarse scale had a good fit to the data as suggested by a large positive correlation between the observed and predicted probabilities of scat counts and the Wilcoxon rank-sum test (z = -1.396, P = 0.163) (Figure 8). 80 Table 17. Estimated coefficients for the AICC selected model (ZINB count) predicting the intensity of activity of latrine sites by river otters, based on coarse-scale habitat data collected on Tezzeron and Pinchi Lakes, central British Columbia from 2007-2008. Parameter Global Model treeheight dspecies(aspen,birch) dspecies(Douglas fir) dspecies (no trees) dspecies(lodgepole pine) dspecies(white spruce) canopy cover dspecies% dbeaverlodge dbeaverlodge2 dreedpatch dreedpatch2 dstreammouth dstreammouth2 waterdepth Constant Inflate Portion treeheight dspecies(aspen,birch) dspecies(Douglas fir) dspeices (no trees) dspecies(lodgepole pine) dspecies(white spruce) canopy cover dspecies% dbeaverlodge dbeaverlodge2 dreedpatch dreedpatch2 dstreammouth dstreammouth2 waterdepth constant Coef. SE 95% CI P 0.024 -0.123 -0.022 0.283 0.176 -0.186 0.003 -0.001 0.049 <-0.001 -0.476 <0.001 -0.339 <0.001 -0.005 2.053 0.013 0.219 0.212 0.610 0.324 0.162 0.004 0.005 0.254 <0.001 0.208 <0.001 0.156 <0.001 0.007 0.485 -0.001 -- 0.050 -0.553 -- -0.307 -0.441 --0.391 -0.912-- 1.480 -0.459 --0.811 -0.250 -- -0.503 -0.005 --0.012 -0.001 --0.001 -0.450 -- 0.548 <-0.001--<0.001 -0.884 -- -0.069 <-0.001--<0.001 -0.644 — -0.0338 <0.001—- <0.001 -0.019-- 0.009 1.101 —-3..01 0.143 0.574 0.918 0.642 0.587 0.250 0.445 0.944 0.847 0.840 0.022 0.090 0.029 0.025 0.482 <0.001 -0.038 0.412 0.422 2.144 -0.546 -0.185 0.009 -0.005 -1.793 <0.001 0.412 <-0.001 0.697 <-0.001 -0.043 -1.866 0.032 0.556 0.922 1.450 0.534 0.458 0.012 0.014 0.821 <0.001 0.556 <0.001 0.499 <0.001 0.463 1.563 -0.100-- 0.024 -0.280 --1.674 -0.012-- 0.036 -0.698 --4.986 -1.593- -0.501 -1.083--0.712 -0.012 —-0.0291 -0.034— <0.023 -3.402- --0.183 <0.001—- <0.001 -0.678 --1.502 <-0.001--<0.001 -0.280 --1.674 <-0.001—- <-0.001 -0.133- - 0.048 -4.929 --1.197 0.229 0.571 0.459 0.139 0.306 0.685 0.820 0.713 0.029 0.014 0.459 0.753 0.162 0.046 0.358 0.233 81 -it—Observed Probability of Count (85%data) -Predicted Probability of Count (15%data) Figure 7. Predicted versus observed probability of scat counts for river otter on Tezzeron and Pinchi Lakes, British Columbia, from May to October (2007-2008). generated with the best fine-scale ZINB model and an independent data set. 82 Predictions were 0.4 H»-~ Observed Probability of Count (85%data) 0.35 -A—Predicted Probability of Count (15%data) 0.3 c 8 0.25 •^ o £• 0.2 * 0.1 0.05 mr ^^m^m^imTmi^atrmr*f^MTmrr*n 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Scat count Figure 8. Predicted versus observed probability of scat counts for river otter on Tezzeron and Pinchi Lakes, British Columbia, from May to October (2007-2008). Predictions were generated with the best coarse-scale ZEMB model and an independent data set. 83 For the residual analysis, the mean difference in the probability of a scat count was closeto zero. The model slightly over-predicted the number of zero counts and to a lesser degree under-predicted the number of single and double scat counts with residuals converging towards zero as the number of scats increased. Discussion I conducted one of the first multi-scale investigations of latrine site use by river otter. The environmental factors dictating animal distribution can vary in importance across scale (Johnson et al. 2002; Forchhammer et al. 2005; Ciarniello et al. 2007); my results provide an additional example of scale-specific selection of resources. Environmental variables had little influence on the distribution of latrine sites at the coarse-scale and activity of otter was most closely associated with features describing the aquatic habitat of fish. In contrast, a number of models explained selection of latrines at the fine scale; both selection and otter activity were influenced by horizontal cover and tree characteristics along the shoreline. A GIS-type analysis of coarse-scale selection would have missed habitat characteristics important to otters within shoreline patches. A multi-scale approach provided a more detailed and complete description and understanding of the selection and use of latrine sites. Multiple spatial and behavioural scales allow us to answer the questions of both what and why otters select for and use latrines. Such inference can provide insights into the possible mechanisms driving the distribution and activity patterns of otters. Only with knowledge of scale specific processes, 84 can we begin to initiate appropriate and efficient conservation and management strategies for otter populations. An ITMC approach was well suited for both the presence/absence and count data collected in this study. Ideally, I would have tested the presence and consistency models on an independent data set, but I lacked the sample size to divide the data into training and testing groups. This was not a result of sampling bias or an insufficient search effort as I located nearly every latrine site (n = 73) across approximately 155 km of shoreline. I did an intensive survey of a large geographic area twice in 2007. Furthermore, results of an intensive survey of 200 random sections of shoreline in 2008 suggested that I was monitoring the majority of latrine sites in the study area during both years. This was also supported by the high percentage (96%) of latrine sites that were identified in 2007 and were also active in 2008. Sample size would have increased only after expanding the study area. I found that habitat characteristics at the fine scale were better at predicting the presence of latrines when compared to coarse-scale environmental features measured at the same sites. The data also suggested that latrine site consistency and intensity of use can be predicted by both fine-scale and coarse-scale habitat characteristics. In general, latrine activity by otters at the coarse scale was best described by aquatic habitat, and by vegetation cover at the fine scale. These results support the hypothesis that otter activity is influenced by habitat characteristics that support their prey at the scale of the landscape, and by habitat characteristics that provide cover at the scale of the shoreline patch. Cover may be important for otters for security from terrestrial predators, to protect scent/scat from the elements and prolong its use for communication, or a combination of both of these functions. 85 At the fine scale, I conducted detailed measurements of latrines that exceeded more simple descriptions of habitat features such as the presence/absence of conifer trees. Using these data, I found that fine-scale variation (e.g., drip-line extent, dbh, number of trees) predicted the presence of latrines and the activity of otters. Selection of latrine sites by otters was positively influenced by large diameter trees and horizontal visual obscurity. Consistently used latrine sites were associated with conifer trees that had a large drip-line extent, a higher number of spruce trees, and increased horizontal cover. Although not significant, the presence and consistency of use of latrine sites was negatively influenced by shrub cover (0-2 m). These results indicate that horizontal cover for otter is important; however, cover is a function of large diameter conifer trees with low hanging branches and not shrubs. Consistent and frequent use of these habitat types suggests they play an especially important role in the ecology of otter populations. Other studies documented the presence of large conifer trees at latrine sites (Newman and Griffin 1994; Swimley et al. 1998), but failed to provide a detailed description of those vegetation communities. For the intensity of activity at latrine sites, shrub cover also had a negative influence on scat numbers; however, unlike the other models, visual obscurity had a significant negative influence on scat numbers. These results suggest that horizontal cover from trees is not as prevalent at latrine sites with a high intensity of use. This difference is most likely explained by the pulsed, frequent visitation rates in areas near the mouth of streams where cover is often not as abundant. This hypothesis is supported by the coarse-scale behavioural models where latrine sites near stream mouths were associated with greater numbers of scats. 86 When tested, the three top-ranked models describing the presence of otter latrines at the landscape scale had poor model accuracy. Otter latrines were well distributed throughout the study area (Figure 1) and the results suggest otters do not select latrine sites based on the coarse-scale variables used in this study, namely food resources and cover. Other studies documented the importance of lakeshore topography, but I did not consider such factors (Newman and Griffin 1994; Swimley et al. 1998). Otters may require an even distribution of latrine sites on the landscape for sociality, or may simply travel frequently and far enough along the shoreline to maintain latrine sites at many locations. A more detailed examination of latrine sites with consistent or high-use visitation by otters produced models that were predictive at the coarse scale. The distance to beaver lodge had a negative influence on activity patterns. This result was at first surprising given previous studies that describe the importance of beaver activity to otters (Melquist and Hornocker 1983; Reid et al 1994a). Reid et al. (1994a), however, found that selection for areas near beaver lodges was at its highest during the winter months. During the winter, lodges are thought to provide cover and access to feeding areas below the ice. In summer, lodges along lakes and streams may not serve as important a role. The relationship between otter distribution and beaver activity in summer may be attributed to the ponds they create rather than the lodges they construct (LeBlanc et al. 2007). The majority of the beaver lodges located in this study were in close proximity to major streams and lakes. Beaver lodges associated with flooded forests, just offshore behind latrine sites, may have not been detected even though they were located a relatively short distance from a latrine site. Lastly, LeBlanc et al. (2007) found that otter activity was most closely associated with beaver ponds 87 that had current resident beavers. No attempt was made to delineate between inactive and active beaver lodges. Exclusion of inactive beaver lodges from the data set may have changed the results. The intensity of activity at latrine sites at the coarse scale was associated with stream mouths and reed patches. Reed patches were also associated with the consistency of activity at latrine sites; these areas may provide important foraging habitat and cover for river otters while hunting fish. In addition, several studies describe otter hunting waterfowl by attacking from underneath while birds float on the water (Meyerriecksl963; Harris 1968; Grenfell 1974). Waterfowl are frequently found in this habitat and the cover provided by reeds may help otters go undetected while hunting. Intensity of activity at stream mouths may be attributed to variable, but high densities of fish during the spawning seasons (e.g., kokanee, sockeye salmon). Consistent with the fluctuations in the availability of this food source, distance to stream mouth was not a significant variable for the consistency model. Activity of otters at the coarse scale likely correlates with the spatial distribution of prey. In addition to fisheries and waterfowl values, management of riparian areas should consider the corequirements of river otter. Considering differences in the ecology of otter populations across their North American range, the results of this study can be generalized in a number of ways. The fineand coarse-scale characteristics of specific habitat features (i.e, conifer drip-line extent or reed patches) that influence otter selection or activity are most applicable to otter populations in central British Columbia. The results of this study, however, may also be relevant to otter populations inhabiting areas with similar predators, prey bases, and habitat types. The 88 specific habitat characteristics may vary (i.e, cover may be provided by different vegetation types or terrain), but the importance of cover at the fine scale and the distribution of food resources at the coarse scale are likely a result of selective pressures common to many river otter populations. Lastly, this study provides an example of the importance and feasibility of combining spatial and behavioral scales into the design of studies that measure wildlife sign. Habitat selection and population monitoring studies that do not include scale in their study design may misinterpret important components of otter habitat or population indices that influence management decisions. For example, important habitat features may be missed when making land use decisions, or wildlife management actions may be misguided by misinterpretations of population trends from surveys of otter sign. Scale is fundamental component in the design and interpretation of ecological investigations. The same ecological processes might show different patterns if observed at a different scale (Wiens 1989; Wheatley and Johnson 2009). The influence of habitat characteristics on latrine site occurrence and activity at one spatial scale may be very different at another scale. The majority of past studies of habitat selection by otters have failed to address this issue aggregating data across scales or focusing on a single scale during study design and analysis (Dubuc et al. 1990; Bowyer et al. 1995; Swimley et al. 1998). Fine-scale measurements of presence/absence data (e.g., presence of conifer trees) were used by Swimley et al. (1998) to measure habitat selection; however, variables at a coarser scale (e.g., minimum distance from coves, tributaries, or islands) were also included. Leblanc et al. (2007) were concerned with selection at the scale of the beaver pond, and Dubuc et al. (1990) focused on selection of habitat by otters at the watershed and forest stand scales. 89 There have been no published studies that investigated latrine selection and activity at multiple spatial and behavioural scales. There are a few examples, however, of past research that has investigated elements of this scalar continuum. For example, Newman and Griffin (1994) adopted a multi-scale approach, relating presence/absence data to habitat characteristics at the fine scale and wetland type categories at the coarse scale. In addition, Newman and Griffin (1994) and Leblanc et al. (2007) investigated otter visitation rates in relation to coarse-scale wetland types. I measured latrine selection and otter behaviour at two spatial scales: landscape and shoreline patch. In addition, this is the first time that consistency and intensity of latrine site activity has been measured in relation to habitat at two spatial scales. Modeling activity patterns at latrine sites, in relation to different spatial and temporal scales, is critical for understanding the mechanisms and processes driving selection and use of latrines by river otters. One reason that past studies of latrine selection have focused on the presence and not the level of activity (measured by scat abundance) at latrine sites is that presence/absence data is much easier to collect (Dubuc et al. 1990; Swimley et al. 1998). My study did not address river otter abundance; however, if deposition rates of scats are directly related to otter abundance, then the number of scats at latrine sites could provide important information on the relative abundance and distribution of otter populations. Nielsen et al. (2005) used two dissimilar species to investigate the relationship between occurrence and abundance. My research provides some tentative support for their conclusions; environmental influences affecting abundance may be different than those limiting distribution. Parameters, coefficients, and models differed when comparing the selection of latrines to measures of 90 otter activity at those sites. This was especially evident when habitat selection and activity were measured at the coarse scale suggesting that different processes may influence the selection and use of latrine sites by otter. These findings, however, are only tentative as I had indices of otter activity only. There has been considerable debate on the utility of using scat surveys to monitor populations of Eurasian otters (Kruuk et al. 1986; MacDonald et al. 1987). In North America, Gallant et al. (2007) cautioned against using latrine sites to predict the number of otter in an area. Their study, however, was conducted during winter when environmental factors influencing otter behaviour and sign are very different than ice-free months. Future studies need to investigate the relationship between otter abundance and the number of scats at latrine sites during the ice-free season. The scales of analysis addressed in my research should be applied to conservation and management actions that incorporate the habitat requirements of otters. The presence of otter latrine sites along shorelines may not be limited by coarse-scale factors such as forest stand type or distance to aquatic features. Habitat characteristics such as conifer trees and cover influence latrine site presence along smaller sections of shoreline, and the level of activity at latrine sites is affected by environmental features at both scales. The interaction among behavioural and spatial scales helps us understand the why behind latrine site selection and use by otters. The more we understand about the why, the more we know about what to protect when managing or conserving otter populations and their habitat. If latrine site consistency and intensity reflect increased activity by otter populations, then habitat characteristics, such as conifer trees and horizontal cover or stream mouths and reed patches, 91 may require additional consideration when prioritizing management actions, protecting areas as a critical habitat, or limiting activities that disturb otter. 92 CHAPTER 4 General Summary 93 General Summary I used an Information Theoretic Model Comparison approach to investigate the relationships among otter diet and temporal/spatial parameters, and habitat characteristics and the presence, consistency, and intensity of otter activity. For river otters in freshwater systems, this study was the first to: 1) model spatio-temporal influences on both the presence of prey items in otter scat and latrine site activity; 2) use stable isotopes to investigate diet; and 3) model latrine selection and activity at multiple spatial and behavioral scales. In Chapter 2, I used a combination of scat inventory and stable-isotope analysis to investigate the contributions of different prey items to the diet of free ranging otter. In general, the stable-isotope analysis agreed with the scat content results showing a dominance of fish in the diet and a small contribution from other prey sources. The stable-isotope analysis, however, suggested a larger contribution from sockeye salmon, larger fish, and birds relative to data from the scat inventory. This study demonstrated another application of stable isotopes in the dietary assessment of a wildlife species. If a small amount of hair or tissue is available, then stable isotopes can provide a more efficient and cost-effective means of investigating otter diets. The specific stable-isotope values and analysis in this study may only be relevant to the central interior of British Columbia, however, the use of this technique could be applied to populations of river otters throughout North America. When modeling the presence of prey items, a combination of fish spawning period, water-body type, and lake best described the presence of salmonids, minnows, and insects in otter scats. The presence of minnows may have also been explained by season and geographic zone. Other food items occurred at low frequencies and could not be modeled, or 94 were modeled and found to be invariant to temporal and spatial parameters. Scat deposition was positively influenced by a time period when no fish were spawning (early July) and by the kokanee spawning period (early September). The influence of kokanee salmon on otter movements was demonstrated in this and previous studies (Melquist and Hornocker 1983). The increase in otter activity at latrine sites in early September was coincident with both the spawning period of kokanee and an increase in the frequency of the Salmonidae family in otter scat. Otters will travel long distances to take advantage of salmonid aggregations (Melquist and Hornocker 1983; Crait and Ben-David 2006). Management agencies could use this information to better guide harvest guidelines for heavily trapped populations. For example, in localized areas where over harvest may be a concern and trapping occurs during spawning seasons, effort could be restricted in stream habitats where otters congregate to feed on salmonid species. In Chapter 3, I investigated the selection of latrine sites and activity of otter at multiple spatial and behavioural scales. This study was one of the first multi-scale investigations of latrine use by river otters. The interactions among spatial and behavioral scales provide a more detailed interpretation of the habitat characteristics influencing latrine use by otters. Relative to the coarse-scale analysis, I found that habitat characteristics at the fine scale were better at predicting latrine sites. In general, fine-scale selection was influenced by parameters that described visual obscurity, larger trees, and characteristics of conifer trees. The presence of latrine sites at the coarse scale was not accurately described by any of the variables I tested. The consistency and intensity of activity at latrine sites at the coarse scale, however, was best predicted by characteristics of aquatic habitat beneficial to 95 fish. These results are a good example of how habitat attributes that are selected at one scale may be missed at another. The use of wildlife sign to measure habitat selection and activity at multiple-scales could have applications to a wide variety of species. Knowledge of the scale to which animals respond to their environment is critical to understanding patterns of behaviour, habitat selection, distribution, and abundance; ecological relationships important for guiding management strategies (Ciarniello et al. 2007). Surveys of animal sign are often employed as an inexpensive means for monitoring trends in wildlife populations. These surveys often provide information about trends in population distribution or abundance, and are used as the basis for management decisions (e.g., Kendall et al. 1992; Patterson et al. 2004). Our ability to monitor the status of animals using sign is limited because we rarely address the assumptions of survey techniques. River otters are difficult to observe in freshwater systems and scat surveys are frequently used to make indirect inferences about otter distribution and relative abundance (Melquist et al. 2003; Pitt et al. 2003; Roberts et al. 2008). The relationships among otter diet, habitat attributes, and the spatio-temporal variation in otter activity at latrine sites helps us understand why river otter are using areas at different times of the year. Monitoring activities that rely on otter sign must be able to differentiate between changes in otter activity that are related to seasonal fluctuations in food resources and/or different habitat types and changes that represent longer-term population trends. Long-term trends in the relative abundance of otter activity may be more apparent if surveys are conducted during the same time period each year to avoid the confounding effects of intra-year variation (Roberts et al. 2008). In addition, the power of sign surveys to reliably index population trends increases with a 96 greater density of sign (Kendall et al. 1992). Concurrent information on diet and activity can provide direction for the timing of surveys to maximize detection of otters and to interpret variations in the relative abundance of otter sign. In addition, spatial and behavioural variations in the use of latrine sites with different habitat attributes can also provide direction for designs of monitoring efforts. For example, surveys for wildlife species often use a random stratified design to maximize detection efforts and increase survey efficiency (Morrison et al. 2001). Habitats that otters use intensively could be given priority in survey designs, and potentially conservation efforts. Although current populations of river otter in central British Columbia appear to be healthy and stable, river otters are sensitive to a range of environmental disturbances. Changes in resource extraction activities or harvest levels could have negative impacts on local populations of river otters and require future conservation and management actions. The information required for management strategies is often gathered as a reaction to low or declining wildlife populations. Information about healthy wildlife populations is critical to the interpretation of population dynamics and the development of recovery plans for threatened populations. This study provides information on the ecology of otters at latrine sites that can be incorporated into survey designs and management strategies of river otter. 97 References Ben-David, M. 1996. Seasonal diets of mink and marten: effects of spatial and temporal changes in resource abundance. Unpublished PhD thesis, University of Alaska. Fairbanks, Alaska. Ben-David, M., G.M. Blundell, J.W. Kern, J.A.K. Maier, E.D. Brown, and S.C. Jewett. 2005. Communication in river otters: Creation of variable resource sheds for terrestrial communities. Ecology 86:1331-1345. Ben-David, M., R.T. Bowyer, L.K. Duffy, D.D. Roby and D.M. Schell. 1998. Social behaviour and ecosystem processes: River otter latrines and nutrient dynamics of terrestrial vegetation. Ecology 79:2567-2571. Ben-David, M., R.W. Flynn, and D.M. Schell. 1997a. Annual and seasonal changes in diets of martens: evidence from stable isotope analysis. Oecologia 111:280-291. Ben-David, M, T.A. Hanley, D.R. Klein, and D.M. Schell. 1997b. Seasonal changes in diets of coastal and riverine mink: the role of spawning Pacific salmon. Canadian Journal of Zoology 75:803-811. Ben-David, M., K. Titus, and L.R. Beier. 2004. Consumption of salmon by Alaskan brown bears: a trade-off between nutritional requirements and the risk of infanticide? Oecologia 13:465-474. Ben-David, M., T.M. Williams, and O.A. Ormseth. 2000. Effects of oiling on exercise physiology and diving behaviour of river otters: a captive study. Canadian Journal of Zoology 78:1380-1390. Blundell, G.M., M. Ben-David, and R.T. Bowyer. 2002. Sociality in river otters: cooperative foraging or reproductive strategies? Behavioural Ecology 13:134-141. Bowyer, T.R., G.M. Blundell, M. Ben-David, S.C. Jewett, T.A. Dean, and L.K. Duffy. 2003. Effects of the Exxon Valdez Oil Spill on River Otters: Injury and Recovery of a Sentinel Species. Wildlife Monographs 153:1-53. Bowyer, R.T., J.W. Testa, and J.B. Faro. 1995. Habitat selection and home ranges of river otters in a marine environment: Effects of the Exxon Valdez oil spill. Journal of Mammology 76:1-11. Burnham, K.P., and D.R. Anderson. 2002. Model selection and inference: a practical information theoretic approach. Second edition. Springer-Verlag, New York, New York. 98 Burnham, K.P., and D.R. Anderson. 2004. Multimodel Inference: Understanding AIC and BIC in model selection. Sociological Methods and Research 33: 261-304. Cannon, D.Y. 1987. Marine Fish Osteology: A Manual for Archeologists. Archeology Press, Department of Archeology, Simon Fraser University. Burnaby, British Columbia. Chatterjee, S., A.S. Hadi, and B. Price. 2000. Regression analysis by example. Wiley. New York. Ciarniello, L.M., M.S. Boyce, D.R. Seip, and D.C. Heard. 2007. Grizzly bear habitat selection is scale dependent. Ecological Applications 17:1424-1440. Costello, CM., and R.W. Sage Jr. 1994. Predicting bear habitat selection from food abundance under 3 forest management systems. International Conference on Bear Resource and Management 9:375-387. Crait, J.R., and M. Ben-David. 2006. River otters in Yellowstone Lake depend on a declining cutthroat trout population. Journal of Mammalogy 87:485-494. Dalerum, F., and A. Angerbjorn. 2005. Resolving temporal variation in vertebrate diets using naturally occurring stable isotopes. Oecologia 144:647-658. Darimont, C.T., and T.E. Reimchen. 2002. Intra-hair isotope analysis implies seasonal shift to salmon in gray wolf diet. Oecologia 80:1638-1642. Dubuc, L.J., W.B. Krohn, and R.B. Owen, Jr. 1990. Predicting occurence of river otters by habitat on Mount Desert Island, Maine. Journal of Wildlife Management 54:594-599. Dodds, R. 2001. Lake re-inventory of Tezzeron Lake for the Upper Fraser Nechako Fisheries Council and Fisheries Renewal BC. Unpublished report. Fort St. James, British Columbia. 9pp. Dunning, J.B., B.J. Danielson, and H.R. Pulliam. 1992. Ecological processes that affect populations in complex landscapes. Oikos 65:169-175. Eide, N.E., J.U. Jepsen, and P. Prestrud. 2004. Spatial organization of reproductive Arctic foxes (Alopex lagopus): responses to changes in spatial and temporal availability of prey. Journal of Animal Ecology. 73:1056-1068. Erlinge, S. 1968. Food studies on captive otters (Lutra lutra). Oikos 19:259-270. 99 Fielding, A.H., and J.F. Bell. 1997. A review of methods for the assessment of prediction error in conservation presence/absence models. Environmental Conservation 24:3849. Forchhammer, M.C., E. Post, T.B.G. Berg, T.T. Hoye, and N.M. Scmidt. 2005. Local-scale and short-term herbivore-plant spatial dynamics reflect influences of large-scale climate. Ecology 86:2644-2651. Foster-Turley, P., S. Macdonald, and C. Mason, (eds), 1990. Otters An Action Plan for their Conservation. IUCN/SSC Otter Specialist Group, Gland, 126 pp. Frair, J.L, E.H. Merrill, D.R. Visscher, D. Fortin, H.L. Beyer, and J.M. Morales. 2005. Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage resources and predation risk. Landscape Ecology 20:273-287. Fritz, H., S. Said and H. Weimerskirch. 2003. Scale-dependent hierarchial adjustments of movement patterns in a long-range foraging seabird. The Royal Society 270:11431148. Fuller, T.K. 1989. Population dynamics of wolves in north-central Minnesota. Wildlife Monographs 105:3-41. Gallant, D., L. Vasseur, and C.H. Berube. 2007. Unveiling the limitations of scat surveys to monitor social species: A case study on river otters. Journal of Wildlife Management 71:258-265. Gard, R. 1961. Effects of the beaver on trout in Sagehen Creek, California. Journal of Wildlife Management 25:221-242. Giere, P., and D.S. Eastman. 2000. American river otters, Lontra Canadensis, and humans: occurrence in a coastal urban habitat and reaction to increased levels of disturbance, pp. 107-125 In H.I. Griffiths (eds.). Mustelids in a modern world. Management and conservation aspects of small carnivore: Human interactions. Backhuys Publishers, The Netherlands. Gilbert, F.F., and E.G. Nancekivell. 1982. Food habits of the mink (Mustela vison) and otter (Lutra canadensis) in northeastern Alberta. Canadian Journal of Zoology 60:12821288. Gonzalez, J.A. 1997. Seasonal Variation in the Foraging Ecology of the Wood Stork in the Southern Llanos of Venezuela. The Condor 99:671-680. Gorman, T.A., J.D. Erb, B.R. McMillan, and D.J. Martin. 2006. Space use and sociality of river otters {Lontra candensis) in Minnesota. Journal of Mammalogy 87:740-747. 100 Greer, K.R. 1955. Yearly food habits of the river otter in the Thompson Lakes Region, northwestern Montana, as indicated by scat analyses. American Midland Naturalist 54:299-313. Grenfell, W.E., Jr. 1974. Food habits of the river otter in Suisin Marsh, central California. Unpublished M.S. Thesis, California State University, Sacramento, California. Harper, R.J., and D. Jenkins. 1982. Moult in the European otter (Lutra lutra). Notes From the Mammal Society 44:298-299. Harris, C.J. 1968. Otters: A study of the Recent Lutrinae. Weidenfield and Nicolson, London, U.K. Hatler, D.F., G. Mowat, and A.M.M. Beal. 2003. Fur Management Guidelines: River Otter (Lontra canadensis). British Columbia Ministry of Envirnoment. British Columbia, Canada. Helon, D.A. 2006. Summer home range, habitat use, movements and activity patterns of river otters (Lontra canadensis) in the Killbuck Watershed, Northeatern Ohio. Unpublished MSc Thesis, West Virginia University. Morgantown, West Virginia. Hendrickx, J. 2001. Contrasts for categorical variables: update. Stata Technical Bulletin 59: 2-5. Hilderbrand, G.V., S.D. Farley, C.T. Robbins, T.A. Hanley, K. Titus and C. Servheen. 1996. Use of stable isotopes to determine diets of living and extinct bears. Canadian Journal of Zoology 74:2080-2088. Hobbs, N.T. 2003. Challenges and opportunities in integrating ecological knowledge across scales. Forest Ecology and Management 181:223-238. Hobson, K.A., B. McLellan, and J. Woods. 2000. Using stable carbon and nitrogen isotopes to infer trophic relationships among black and grizzly bears in the Upper Columbia River basin, British Columbia. Canadian Journal of Zoology 78:1332-1339. Hosmer, D.W., and S. Lemeshow. 2000. Applied Logistic Regression, 2nd Edition. John Wiley and Sons. New York. Johnson, C.J., K.L. Parker, D.C. Heard, and M.P. Gillingham. 2002. Movement parameters of ungulates and scale-specific responses to the environment. Journal of Animal Ecology 71:225-235. 101 Johnson, C.J., K.L. Parker, and D.C. Heard. 2001. Foraging across a variable landscape: behavioral decisions made by woodland caribou at multiple spatial scales. Oecologia 127:590-602. Johnson, D.H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71. Johnston, N.T., J.S. Macdonald, K.J. Hall, and P.J. Tschaplinski. 1997. A preliminary study of the role of Sockeye Salmon (Onchorhynchus nerka) carcasses as carbon and nitrogen sources for benthic insects and fishes in the "Early Stuart" stock spawning streams, 1050 km from the ocean. Fisheries Project Report No. RD55. Province of British Columbia, Ministry of Environment, Lands and Parks, Fisheries Branch. Kalz, B., K. Jewgenow, and J. Fickel. 2006. Structure of an otter (Lutra lutra) population in Germany- results of DNA and hormone analyses from faecal samples. Mammalian Biology 71:321-335. Kendall, K.C., L.H. Metzgar, D.A. Patterson, and B.M. Steele. 1992. Power of sign surveys to monitor population trends. Ecological Applications 2:422-430. Kimber, K.R., and G.V. Kollias. 2000. Infectious and parasitic diseases and contaminant related problems of North American river otters (Lontra canadensis): A review. Journal of Zoo and Wildlife Medicine 31:452-472. Kruuk, H. 1992. Scent marking by otters (Lutra lutra): signaling the use of resources. Behavioural Ecology 3:133-140. Kruuk, H., J.W.H. Conroy, U. Glimmerveen, and E.J. Ouwerkerk. 1986. The use of spraints to survey populations of otters Lutra lutra. Biological Conservation 35:187-194. Lagler, K.F. 1970. Freshwater fishery biology. W.C. Brown Co., Dubuque, Iowa. Larsen, D.N. 1984. Feeding habits of river otters in coastal southeastern Alaska. Journal of Wildlife Management 48:1446-1452. LeBlanc, F.A., Gallant, D., Vasseur, L., and L. Leger. 2007. Unequal summer use of beaver ponds by river otters: influence of beaver activity, pond size, and vegetation cover. Canadian Journal of Zoology 85:774-782. Long, J.S., and J. Freese. 2006. Regression Models for Categorically Dependent Variables Using Stata. 2nd Ed. Stata Press. College Station, Texas. Lukacs, P.M., and K.P. Burnham. 2005. Review of capture-recapture methods applicable to noninvasive genetic sampling. Molecular Ecology 14: 3909-3919. 102 Macdonald, S.M., and C.F. Mason. 1987. Seasonal marking in an otter population. Acta Theriologica 32:449-462. Maurel, D., C. Coutant, L. Boissin-Agasse, and J. Boissin. 1986. Seasonal moulting patterns in three fur bearing mammals: the European badger (Meles meles L.), the red fox (Vulpes vulpes L.), and the mink (Mustela vison). A morphological and histological study. Canadian Journal of Zoology 64:1757-1764. McAllister, D.E., and C.C. Lindsey. 1961. Systematics of the Freshwater Sclupins (Cottus) of British Columbia. Bulletin of the National Museum of Canada, Contributions to Zoology. No. 172 (1959):66-89. McDowell, D.M., and R.J. Naiman. 1986. Structure and function of a benthic invertebrate stream community as influenced by beaver (Castor canadensis). Oecologica 68:481489. McPhail, J.D. 2007. The freshwater fishes of British Columbia. The University of Alberta Press. Edmonton, Alberta. Melquist, W.E., and M.G. Hornocker. 1983. Ecology of river otters in west central Idaho. Wildlife Monographs 83:1-60. Melquist, W.E., P.J. Polechla, Jr., and D. Toweill. 2003. River Otter, pp. 708-734 In G.A. Felhamer, B.C. Thompson, and J.A. Chapman (eds.). Wild mammals of North America: management, and conservation. John Hopkins University Press, Baltimore, Maryland. Menard, S. 2001. Applied Logistic Regression Analysis. Sage University Paper Series on Qunatitative Applications in the Social Sciences, 07-106. Thousand Oaks, California: Sage. Meyerriecks, A.J. 1963. Florida otter preys on common gallinule. Journal of Mammal Research 44:425-426. Milakovic, B. 2008. Defining the predator landscape of northeastern British Columbia. Unpublished Ph.D. Thesis, University of Northern British Columbia. Prince George, British Columbia. Molsher, R.L., E.J. Gifford, and J.C. Mcllroy. 2000. Temporal, spatial, and individual variation in the diet of red foxes (Vulpes vulpes) in central New South Wales. Wildlife Research 27:593-601. 103 Morrison, M.L., W.M. Block, M.D. Strickland, W.L. Kendall. 2001. Wildlife Study Design. Springer-Verlag, New York Inc. New York. Mowat, G., and D.C. Heard. 2006. Major components of grizzly bear diet across North America. Canadian Journal of Zoology 84:473-489. Nelson, J.S. 1973. Morphological differences between the Teleosts Couesius plumbeus (Lake Chub) and Rhinichthys cataractae (Longnose Dace) and their hybrids from Alberta. Journal of Morphology 139:227-237. Newman, D.G., and C.R Griffin. 1994. Wetland use of river otters in Massachusetts. Journal of Wildlife Management 58:18-23. Nielsen, S.E., Johnson, C.J., Heard, D.C, M.S. Boyce. 2005. Can models of presenceabsence be used to scale abundance? Two case studies considering extremes in life history. Ecography 28:197-208. Patterson, B.R., N.W.S. Quinn, E.F. Becker, D.B. Meier. 2004. Estimating Wolf Densities in Forested Areas Using Network Sampling of Tracks in Snow. Wildlife Society Bulletin 32:938-947. Pearce, J., and S. Ferrier. 2000. Evaluating the predictive performance of habitat models developed using logistic regression. Ecological Modeling 133:225-245. Pitt, J.A., W.R. Clark, R.D., R.D. Andrews, K.P. Schlarbaum, D.D. Hoffman, and S.W. Pitt. 2003. Restoration and monitoring of river otter populations in Iowa. Journal of the Iowa Academy of Sciences 11:7-12. Phillips, D.L., and J.W. Gregg. 2001. Uncertainty in source partitioning using stable isotopes. Oecologica 127:171-179. Phillips, D.L., and J.W. Gregg. 2003. Source partitioning using stable isotopes: coping with too many sources. Oecologia 136:261-269. Phillips, D.L., S.D. Newsome, and J.W. Gregg. 2005. Combining sources in stable isotope mixing models: alternative methods. Oecologia 144:520-527. Prigioni, C, L. Remonti, A. Balestrieri, S. Sgrosso, G. Priore, N. Mucci, and E. Randi. 2006. Estimation of European otter (Lutra lutra) population size by fecal DNA typing in southern Italy. Journal of Mammalogy 87:855-858. Quinn T.P., S.M. Carlson, S.M. Gende, and H.B. Rich, Jr. 2009. Transportation of Pacific salmon carcasses from streams to riparian forests by bears. Canadian Journal of Zoology 87:195-203. 104 Raesly, E.J. 2001. Progress and status of river otter reintroduction in the United States. Wildlife Society Bulletin 29:856-862. Reid, D.G., T.E. Code, A.C.H. Reid, and S.M. Herrero. 1994a. Spacing, movements, and habitat selection of the river otter in boreal Alberta. Canadian Journal of Zoology 72:1314-1324. Reid, D.G., T.E. Code, A.C.H. Reid, and S.M. Herrero. 1994b. Food habits of the river otter in a boreal ecosystem. Canadian Journal of Zoology 72:1306-1313. Resources Inventory Committee. 2006. Data collection standards for Vegetation Resource Inventory Ground Sampling. Forest Analysis and Inventory Branch, Ministry of Forests and Range, Victoria, British Columbia. Resources Inventory Committee. 2008. Vegetation Resource Inventory Data Dictionary. Forest Analysis and Inventory Branch, Ministry of Forests and Range, Victoria, British Columbia. Roberts, N.M., S.M. Crimmins, D.A. Hamilton, and E. Gallagher. 2008. An Evaluation of Bridge-sign Surveys to Monitor River Otter (Lontra canadensis) Populations. The American Midland Naturalist 160:358-363. Rostain, R.R., M. Ben-David, P. Groves, and J.A. Randall. 2004. Why do river otters scentmark? An experimental test of several hypotheses. Animal Behaviour 68:703-711. Roth, J.D. 2002. Temporal variability in arctic fox diet as reflected in stable-carbon isotopes; the importance of sea ice. Oecologia 133:70-77. Stenson, G.B. 1985. The reproductive cycle of the river otter, Lutra canadensis, in the marine environment of southwestern British Columbia. Unpublished PhD Thesis, University of British Columbia. Vancouver, British Columbia. Stenson, G.B., G.A. Badgero, and H.D. Fisher. 1984. Food habits of the river otter Lutra canadensis in the marine environment of British Columbia. Canadian Journal of Zoology 62:88-91. Sulkava, R. 2006. Ecology of the otter {Lutra lutra) in central Finland and methods for estimating the densities of populations. Unpublished PhD Thesis, University of Joensuu. Joensuu, Finland. Swets, J.A. 1988. Measuring the accuracy of diagnostic systems. Science 240:1285-1293. 105 Swimley, T.J., T.L. Serfass, R.P. Brooks, and W.M. Tzilkowski. 1998. Predicting river otter latrine sites in Pennsylvania. Wildlife Society Bulletin 26:836-845. Tiezen, L.L., T.W. Boutton, K.G. Tesdahl., and N.A. Slade. 1983. Fractionation and turnover of stable carbon isotopes in animal tissues: implications for 813C analysis of diet. Oecologica 57:32-37. Toweill, D.E. 1974. Winter food habits of river otters in western Oregon. Journal of Wildlife Management 3 8:107-111. Urton, E.J.M., and K.A. Hobson. 2005. Intrapopulation variation in gray wolf isotope profiles: implications for the ecology of individuals. Oecologia 145:317-326. Vuong, Q.H. 1989. Likelihood ratio tests for model selection and non-tested hypotheses. Econometrica 57:307-333. Wheatley, M., and C. Johnson. 2009. Factors limiting our understanding of ecological scale. Ecological Complexity 6:150-159. Wiens, J.A. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397. Wilson, K.A. 1954. The role of mink and otter as muskrat predators in northeastern North Carolina. Journal of Wildlife Management 18:199-207. 106 Appendix 1. Stable isotopic signature means, standard error, and 95% confidence intervals for potential prey items of river otter (Lontra canadensis) on Tezzeron and Pinchi Lakes, central British Columbia (2007-2008). SEC Sample Prey Group SN SEN CI+ CI5C CI+ CIsize Bird (red-necked 0.58 5.83 -29.81 0.04 -29.73 -29.89 4 6.96 8.09 grebe, teal species) Clam 2.41 -31.87 0.01 -31.84 -31.90 4 2.47 0.03 2.52 Insect (caddis fly, 3.70 0.67 5.02 2.39 -29.26 1.15 -27.00 -31.52 5 may fly) Mammal (Beaver, 2.77 0.98 4.69 0.84 -24.14 0.92 -22.33 -25.95 4 Muskrat) Otter 22 10.61 0.23 11.05 10.16 -26.61 0.39 -25.84 -27.38 11.27 0.47 -25.84 -27.68 3 Burbot 11.37 0.05 11.46 -26.76 8.93 -33.00 0.21 -32.58 -33.42 3 Kokanee 9.30 0.19 9.66 7.22 3 Lake chub 0.11 7.44 7.01 -26.50 0.98 -24.57 -28.42 11.74 0.54 Lake trout 3 12.79 10.69 -29.73 1.50 -26.79 -32.67 Lake trout/burbot 6 11.59 0.25 12.08 11.10 -28.76 0.80 -27.20 -30.33 3 Northern 8.69 0.37 9.42 7.96 -27.69 0.52 -26.68 -28.70 pikeminnow Rainbow trout 7.85 0.20 8.25 7.45 -27.28 1.21 -24.91 -29.65 4 Sculpin species 0.12 6.34 5.87 -26.95 0.45 -26.07 -27.84 3 6.10 4 Sockeye salmon 11.13 0.08 11.29 10.96 -22.55 0.37 -21.82 -23.27 Sucker species 0.71 6.19 -27.20 0.44 -26.35 -28.06 3 7.57 8.96 Whitefish 0.10 7.22 -31.04 0.19 -30.67 -31.41 3 7.40 7.59 24 Aggregate fish 1 * 0.33 9.12 7.85 -28.60 0.43 -27.77 -29.44 8.48 0.19 7.16 -27.96 0.42 -27.14 -28.79 15 Aggregate fish 2* 7.53 7.90 *Fish 1: rainbow trout, lake trout, kokanee, sucker species, northern pikeminnow, whitefish, burbot, and sculpin species. **Fish 2: rainbow trout, sucker species, northern pikeminnow, whitefish, and sculpin species. 107 Appendix 2. Detailed description of fine-scale habitat measurements at river otter latrine sites on Tezzeron and Pinchi Lakes, central British Columbia (2007-2008). 1) Visual obscurity: Cover pole placed 1 m inland and perpendicular to plot center. Observer stands 5 m from cover pole in direction of water and perpendicular to shoreline (% obscured at three heights (0-.5 m, .5-1.0 m, 1.0 m-1.5 m) rounded to the nearest 5%). 2) % Cover: Estimated % cover to the nearest 5% using an ocular estimate. Tree layer includes all woody plants >10 m tall; shrub layer includes tall shrubs (2-10 m tall, Bl) and low shrubs (<2 m tall, B2); Herb layer includes all herbaceous species regardless of height and woody plants <15 cm tall. 3) Conifer cover: Includes only trees in which the trunk originates within the plot boundaries. Measure distance from trunk to drip line extent. 4) Bank height: Measured distance from base of bank to top of bank in centimeters located between plot center and water line. 5) Slope: Percent slope measured within latrine site using a clinometer. 6) Substrate: Record substrate between water and origin of terrestrial vegetation (Mud, Sand, Cobble, Boulder). 7) Tree species and characteristics: Includes only trees in which trunk originates within the plot boundaries. Recorded data included: species, diameter at breast height (dbh) on trees >7.5 cm, and distance from vegetation line to tree trunk. 108