Tamiasciurus Hudsonicus) in Highly Fragmented Secondary Hardwood Forests

FOREST FRAGMENTATION EFFECTS AND THE CAVITY NEST MATERIAL REQUIREMENTS OF

NORTHERN FLYING SQUIRRELS AND RED SQUIRRELS IN A FRAGMENTED SECONDARY

HARDWOOD FOREST REGION OF ONTARIO, CANADA

Jesse Eric-Henry Patterson

A thesis submitted in conformity with the requirements

for the degree of Master of Science in Forestry

Faculty of Forestry

University of Toronto

395 Wellington Street 395, rue Wellington Ottawa ON K1A0N4 Ottawa ON K1A0N4 Canada Canada

Your file Votre reference ISBN: 978-0-494-39741-1 Our file Notre reference ISBN: 978-0-494-39741-1

NOTICE: AVIS: The author has granted a non L'auteur a accorde une licence non exclusive exclusive license allowing Library permettant a la Bibliotheque et Archives and Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par Plntemet, prefer, telecommunication or on the Internet, distribuer et vendre des theses partout dans loan, distribute and sell theses le monde, a des fins commerciales ou autres, worldwide, for commercial or non sur support microforme, papier, electronique commercial purposes, in microform, et/ou autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in et des droits moraux qui protege cette these. this thesis. Neither the thesis Ni la these ni des extraits substantiels de nor substantial extracts from it celle-ci ne doivent etre imprimes ou autrement may be printed or otherwise reproduits sans son autorisation. reproduced without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne Privacy Act some supporting sur la protection de la vie privee, forms may have been removed quelques formulaires secondaires from this thesis. ont ete enleves de cette these.

While these forms may be included Bien que ces formulaires in the document page count, aient inclus dans la pagination, their removal does not represent il n'y aura aucun contenu manquant. any loss of content from the thesis. Canada Patterson, Jesse Eric-Henry. 2008. Forest fragmentation effects and the cavity nest material requirements of northern flying squirrels and red squirrels in a fragmented secondary hardwood forest region of Ontario, Canada. Master of Science in Forestry, Faculty of Forestry, University of Toronto.

ABSTRACT

Agriculturally induced fragmentation of hardwood forests has been extensive in southern Ontario, with important implications for biodiversity, such as the outright loss of species, contraction in species distributions, and reduced genetic diversity. The purpose of this thesis was to investigate the patterns of fragment occupancy, from a metapopulation perspective, as a function of landscape-level features and to identify significant aspects of habitat use, including nest material requirements, for northern flying squirrels {Glaucomys sabrinus) and red squirrels (Tamiasciurus hudsonicus) in highly fragmented secondary hardwood forests. The occurrence of G. sabrinus was positively correlated with fragment area, while T. hudsonicus occurrence was positively correlated with the basal area of coniferous trees. Both species primarily used shredded eastern white cedar {Thuja occidentalis) bark as nesting material, possibly due to a behavioural adaptation to reduce ectoparasite loads in the nest environment.

II ACKNOWLEDGMENTS

First, and foremost, I would like to thank my supervisor, Jay Malcolm, for his consistent

intellectual support, advice, guidance, encouragement, enthusiasm, and master computer

programming!

My supervisory committee is gratefully acknowledged for their input and interest: Jeff

Bowman, Danijela Puric-Mladenovic, and Sandy Smith. Their perspectives and insights significantly improved this thesis.

I would like to thank the Grey Sauble Conservation Authority (GSCA), the Saugeen Valley

Conservation Authority (SVCA) and the various private landowners who graciously made their land available for this study. I am also indebted to the GCSA, SVCA, and Ontario Ministry of Natural

Resources (OMNR) for providing necessary landscape imagery, in particular Chris Hachey (GSCA),

Jim Penner (SVCA), and Danijela Puric-Mladenovic (OMNR).

Thistlewood Timber Frame Homes in Markdale, Ontario generously donated the lumber used for constructing the nest boxes, and for that I am grateful. The Patterson and Sharkey families are both owed a great debt of gratitude for graciously providing in-kind accommodation throughout the course of this study.

Without the help of my field assistants this project would have never come to fruition. I would like to acknowledge Carly Armstrong, Stephen Patterson, Marianne Patterson, and Hillary

Maddin for their field, logistic, and technical support during data collection. I also thank the Wildlife

Ecology lab and my family and friends for two-years of discussion, support, and friendship.

Funding for this project was provided by a Natural Sciences and Engineering Research

Council (NSERC) Canada Graduate Scholarship to JEHP, an NSERC Undergraduate Summer

Research Award to JEHP, and an NSERC research grant to J. Malcolm.

iii TABLE OF CONTENTS

ABSTRACT ii

ACKNOWLEDGMENTS iii

TABLE OF CONTENTS iv

LISTOFTABLES vi

LIST OF FIGURES vii

GENERAL INTRODUCTION 1

STUDY AREA DESCRIPTION 5

CHAPTER 1: RELATIVE INFLUENCE OF LANDSCAPE STRUCTURE AND FRAGMENT AREA ON PATTERNS OF NORTHERN FLYING SQUIRREL AND RED SQUIRREL OCCURRENCE IN A SECONDARY HARDWOOD FOREST

ABSTRACT 7

1. INTRODUCTION 8

2. METHODS 15

2.1 Site Selection 15 2.2 Sciurid Trapping 16 2.3 Habitat Measurements 18 2.4 Landscape Measurements 19

2.4.1 Mean Proximity Index 21 2.4.2 Mean Nearest Neighbour Distance 22 2.4.3 Proportion of Forest 23 2.4.4 Number of Patches 23

2.5 Statistical Analysis 23

2.5.1 Landscape Scale 23

2.5.2 Bivariate and multivariate approaches 25

3. RESULTS 27

3.1 Sciurid Abundance and Occurrence 28

iv 3.2 Spatial Independence 28 3.3 Landscape Scale 28 3.4 Logistic Regression 31 3.5 Redundancy Analysis 40

4. DISCUSSION 43

4.1 Scale of best model fit 43 4.2 Responses of Glaucomys sabrinus to habitat features 46 4.3 Responses of Glaucomys sabrinus to landscape structure features 50 4.4 Responses of Tamiasciurus hudsonicus to habitat features 54 4.5 Responses of Glaucomys sabrinus to landscape structure features 55 4.6 Misclassifying absence 57 4.7 Future directions and conclusions 58

CHAPTER 2: CAVITY NEST MATERIAL USE BY NORTHERN FLYING SQUIRRELS AND RED SQUIRRELS IN SOUTHERN ONTARIO: A CASE FOR THE NEST-PROTECTION HYPOTHESIS

ABSTRACT 61

1. INTRODUCTION 62

2. METHODS 64

3. RESULTS 66

3.1 Nest Box Occupation 66 3.2 Nest Materials of Glaucomys sabrinus and Tamiasciurus hudsonicus 67

3.3 Nest Depth 69

4. DISCUSSION 71

GENERAL CONCLUSIONS 76

LITERATURE CITED 80

v LIST OF TABLES

Table 1.1: Landscape and habitat variables for Glaucomys sabrinus in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region. Slope, Pr(>lzl) and corrected Akaike Information Criterion (AICc) values are from logistic regression analysis for each individual predictor variable. Scales of best model fit and transformations are given in brackets beside corresponding terms. Bold indicates significant values (p < 0.05). Mean values (± 1 SD) are given separately for sites where Glaucomys sabrinus was present and absent 32

Table 1.2: As Table 1.1 except that data are for Tamiasciurus hudsonicus 33

Table 1.3: Effect of the most significant habitat variables on predicting Glaucomys sabrinus occurrence in secondary hardwood Great Lakes-St. Lawrence forests with the inclusion of the most significant landscape variables forced into a multiple logistic regression model. Significance level for entry into the model was a = 0.1; for exit it was 0.15. Characteristic scales of response and transformations are given in brackets beside corresponding terms 35

Table 1.4: As Table 1.3 except that data are for Tamiasciurus hudsonicus 36

Table 1.5: Results of redundancy analyses evaluating the significance of each landscape variable on predicting Glaucomys sabrinus occurrence without and with patch area partialled out (i.e., area entered as a covariate). Monte Carlo statistical tests were run with 9999 unrestricted permutations and species scores were centred and standardised 44

vi LIST OF FIGURES

Figure 1.2: Corrected Akaike Information Criterion (AICc) values from logistic regressions between Glaucomys sabrinus occurrence and mean proximity index, mean nearest neighbour distance, proportion of forest, and number of patches as a function of landscape radius for spatially independent (18 sites, lower line plot) and non-spatially independent (24 sites, upper line plot) data sets. The radius of best model fit is indicated (drop-down arrow) and Spearman correlation results (p and rs values) between the spatially independent and non-spatially independent AICc values are given 29

Figure 1.3: As Figure 1.2, except that values are for Tamiasciurus hudsonicus 30

Figure 1.4: Presence/absence of Glaucomys sabrinus in relation to the logio-transformed patch area in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region, with a fitted probability of response curve derived from the logistic regression model (p = 0.016, AICc = 21.5) 37

Figure 1.5: As Figure 1.4, except that presence/absence of Tamiasciurus hudsonicus is shown in relation to the basal area of deciduous trees in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region (p = 0.047, AICc = 31.94) 38

Figure 1.6: Receiver operating characteristic (ROC) curve based on a logistic regression model for the presence/absence of Glaucomys sabrinus as a function of the probability of Glaucomys sabrinus occurrence given logio patch area in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region 39

Figure 1.7: Principal component analysis of all independent landscape variables at their respective scales of best model fit showing sites where Glaucomys sabrinus was present and absent in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region. Inter-sample scaling was used and species scores were centred and standardised 41

Figure 1.8: As Figure 1.7, except for Tamiasciurus hudsonicus 42

vii likely to occur. Two Universal Transverse Mercator, Zone 17 grid lines also are shown (map datum: NAD83) 51

Figure 2.1: Total number of cavity nests from the complete sample (n = 224) (i.e., known and unknown occupants) dominated (> 50% by volume) by the various materials in a secondary hardwood Great Lakes-St. Lawrence forest region 68

Figure 2.2: Species-specific distribution of dominant (> 50%) nest materials found in nest boxes known to be occupied by Glaucomys sabrinus (n = 14) and Tamiasciurus hudsonicus (n= 18) in a secondary hardwood Great Lakes-St. Lawrence forest region 70

viii GENERAL INTRODUCTION

Small mammals play important ecological roles in forest ecosystems, acting as prey, predators, and seed and spore dispersers, and are a significant component of biodiversity in their own right. As forest fragmentation by agriculture, urbanisation, and forestry intensifies, understanding the biological mechanisms underlying small mammal responses to fragmentation will become increasingly important to biodiversity conservation and resource management. Habitat size, connectedness, and quality all assume important roles in species preservation (Weigl 2007) and all of these are affected by fragmentation. Fragmentation results in outright habitat loss; it also results in the creation of barriers to animal movement that are often permanent in nature (e.g., roads, agriculture) and in alterations of habitat size and quality. Fragmentation of small mammal populations can destroy metapopulation structured groups with habitat isolates becoming more susceptible to local extinction and little to no patch recolonisation.

Arboreal squirrels (Rodentia: Sciuridae) are forest obligate species and appear to provide excellent models for which to assess the influence of forest fragmentation on distribution patterns (Koprowski 2005). Habitat (forest) fragments are relatively easily defined for arboreal squirrels and the ease of trapping tree squirrels appears to allow for rather straightforward determination of species occurrence in forest fragments (Koprowski

2005). Two such tree squirrels, the northern flying squirrel {Glaucomys sabrinus Shaw) and the red squirrel {Tamiasciurus hudsonicus Erxleben), appear to be thriving in the northern and western portions of their respective ranges; however, the future of both species is uncertain in southerly parts of their range, such as southern Ontario, where substantial

1 2 portions of the original forest cover have been converted to non-forested land covers. For instance, in Bruce and Grey counties, Ontario, the forests that once dominated over 90% of the landscape in presettlement times now only cover approximately 20% of the landscape

(Riley and Mohr 1994). The remaining forest tracts are often small, isolated, and heavily managed (Pearce 1992). Small forest fragments generally preclude the occurrence of forest interior and forest-obligate species (Friesen et al. 1999), however, the implications of shrinking patch area are unknown for many species. Presumably, the reduced area and increased isolation of habitat patches over time has produced declines in population abundances and contractions of species ranges beyond what might be expected from habitat loss alone (Andren 1994). Only through a thorough understanding of individual species biology and, in particular, their responses to fragmentation can we identify practical and effective means to mitigate cumulative negative effects of forest modification and habitat fragmentation.

In broad terms, G. sabrinus is a noctural, nonhibernating, relatively long-lived (>7 years; Villa et al. 1999) gliding tree squirrel common to boreal and northern temperate forests throughout Canada and parts of the United States. Glaucomys sabrinus is commonly considered to be a habitat specialist in that it has been found to achieve maximum densities in forests with abundant downed woody debris, snags, and tall trees (Carey et al. 1999,

Rosenberg and Anthony 1992, Witt 1992, Holloway and Malcolm 2006). However, more recent literature suggests that the behaviour of G. sabrinus may be quite plastic and that this sciurid is capable of thriving in a wide range of habitat conditions (Weigl 2007). Although G. sabrinus may be a habitat opportunist, habitat preservation and management nonetheless remain major conservation issues for this forest species. Glaucomys sabrinus may be 3 particularly sensitive to changes in landscape structure because of its gliding locomotion, which may be poorly suited for moving across non-forested habitats. Hence, G. sabrinus dispersal may be especially sensitive to forest fragmentation, with small subpopulations more vulnerable to local extinction and increased inbreeding depression. Given their group huddling life-history strategy to conserve energy in winter months (Wells-Gosling and

Heaney 1984), G. sabrinus may be at an elevated risk of subpopulation extinction in small habitat isolates since smaller populations may not be able to satisfy individual thermoregulatory demands during the winter. As of yet, few studies have examined G. sabrinus populations in fragmented landscapes. The two such studies that have been conducted focussed nearly exclusively on patch-area relationships in boreal forest regions with > 30% of the original habitat remaining (Bayne and Hobson 1998, Smith and Person

2007). Much needed are studies of G. sabrinus response to isolation, connectivity, and proportion of nearby habitat in regions with < 30% remaining forest cover (at appropriate scales), as well as G. sabrinus dispersal abilities and use of matrix.

Tamiasciurus hudsonicus is a spruce (Picea sp.) specialist and conifer-obligate throughout most of its range in North America; however, it does occupy deciduous- dominated forests in the more eastern and southern portions of its range. In like manner to G. sabrinus, T. hudsonicus is a secondary cavity nester and is widely considered to rely on snags and cavity trees for natal and winter nesting (Layne 1954, Rothwell 1979). In central Ontario,

T. hudsoncicus is associated with a high density of large spruce and hardwood trees, snags in conifer sites, and low levels of logging (Holloway and Malcolm 2006). Studies of T. hudsonicus responses to habitat fragmentation have produced conflicting results, perhaps as a function of differences in landscape context and habitat quality (Bayne and Hobson 2000, 4

Nupp and Swihart 2000, Goheen et al. 2003a). Yet, in deciduous-dominated forest regions throughout the midwestern United States, T. hudsonicus has successfully managed to expand its range concurrent with agriculturally-induced forest fragmentation and afforestation of coniferous trees, presumably due to its good vagility and high population growth potential

(Swihart and Nupp 1998).

Cavity nest materials and nest construction methods employed by G. sabrinus and T. hudsonicus have received little research focus in the scientific literature (except see Hayward and Rosentreter 1994). Nests provide small mammals with sites for raising young, avoiding predators, and improving thermoregulation (Weigl 1978). The materials used by small mammals to line their nests often act to further improve thermoregulation (Stapp et al. 1991), provide additional food sources (Hayward and Rosentreter 1994), and reduce ectoparasite loads in the nest environment (Hemmes et al. 2002). Suitable nesting substrates are likely an important resource for both G. sabrinus and T. hudsonicus and their availability may be a potential ecological factor influencing these species distributions and densities in secondary hardwood Great Lakes-St. Lawrence forests. In order to better conserve G. sabrinus and T. hudsonicus, land managers must be aware of their regionally-specific habitat requirements.

In Chapter 1 of this thesis I explore the habitat relationships of G. sabrinus and T. hudsonicus in deciduous-dominated, second growth Great Lakes-St. Lawrence forests

(Huron-Ontario Region; Rowe 1972) and relate G. sabrinus and T. hudsonicus patch occupancy to parameters of the landscape, including measurements of patch area and isolation, and to habitat availability. In Chapter 2,1 characterise the cavity nest materials and cavity nest construction for G. sabrinus and T. hudsonicus in a Great Lakes-St. Lawrence forest, focussing on the importance of eastern white cedar bark as a nesting resource. 5

STUDY AREA DESCRIPTION

The study was conducted in Bruce and Grey counties, Ontario, Canada (43°59' to

45°12'N, 80°22' to 81°39'W), which together encompass an area of about 22000 km2. The climate in the region is humid continental and heavily influenced by Lake Huron to the west and Georgian Bay to the east (mean annual temperature: 6.1°C; mean annual precipitation:

1041 mm). The majority of precipitation falls as rain (740 mm), with an additional 426 mm of snow falling annually from November to March. Average winter snow depth is 21 cm and, on average, 138 days per annum have a maximum daily temperature below 0°C.

Virgin, Great Lakes-St. Lawrence forest dominated Bruce and Grey counties until the late 1800s (Rowe 1972), at which time extensive logging, forest fires, and agricultural settlement eliminated most of the primary, closed-canopy forest (De Gruchy et al. 2005).

Currently, the main contributors to forest disturbance are row-crop agricultural, small-scale forestry, roads, and commercial and residential development. Forest fragmentation is severe in Bruce and Grey counties, which is one of the most intensively farmed regions in Canada

(Allen et al. 1990). Today, roughly 22% of the landscape is forested and most tracts of forest habitat are smaller than 100 ha (Friesen et al. 1999). Closed-canopy forest in Bruce and Grey counties is dominated by sugar maple (Acer saccharum Marsh), white ash (Fraxinus americana L.), American beech (Fagus grandifolia Ehrh), eastern white cedar (Thuja occidentalis L.), and aspen (Populus tremuloides Michx, P. grandidentata Michx), and white birch (Betula papyrifera Marshall) (Rowe 1972). Small stands of eastern white pine (Pinus strobus L.) and spruce (Picea glauca Voss, P. mariana Mill.) occur throughout the area, often in plantations. White ash is known to be a common associate of remnant deciduous fragments in agricultural landscapes throughout southern Ontario (Pedlar et al. 1997). Eastern hemlock (Tsuga canadensis [L.] Carr.), a historically common tree species, has been all but eliminated in Bruce and Grey counties over the past century (Rowe 1972). CHAPTER 1: RELATIVE INFLUENCE OF LANDSCAPE STRUCTURE AND FRAGMENT AREA ON

PATTERNS OF NORTHERN FLYING SQUIRREL AND RED SQUIRREL OCCURRENCE IN A

SECONDARY HARDWOOD FOREST

ABSTRACT

In this study I investigate patterns of northern flying squirrel (Glaucomys sabrinus) and red squirrel (Tamiasciurus hudsonicus) occurrence in forest fragments and relate those patterns to habitat and landscape structure features in a fragmented deciduous-dominated landscape.

Presence-absence of G. sabrinus and T. hudsonicus was assessed in 24 forest fragments of various sizes through live-trapping methods. By use of a geographic information system, I calculated five metrics to quantify the patch context of each focal habitat fragment at scales >1 km: landscape composition (number of patches, proportion of forest), landscape configuration

(mean proximity index, mean nearest neighbour distance), and focal fragment area. Appropriate scales of study were determined by measuring the best model fit of G. sabrinus and T. hudsonicus response to the patch context variables at ten spatial scales (1000 m to 10000 m) by use of logistic regression. I found that G. sabrinus occurrence was positively correlated with fragment size (/?=0.016) and speculate that G. sabrinus subpopulations either abandon or experience local extinctions in small habitat patches more frequently than large patches, perhaps due to thermoregulatory pressures. G. sabrinus did not appear to respond to any of the measured habitat variables, whereas the occurrence of T. hudsonicus was highly associated with increased basal area of coniferous trees (p=0.047). Unlike G. sabrinus, T. hudsonicus did not appear to respond to differences in landscape structure, perhaps due to greater vagility and stronger habitat affinities.

7 8

1. INTRODUCTION

Habitat fragmentation, that is, the formation of isolated fragments of habitat from a once continuous extent (Harris 1984), occurs naturally, for example, through fire and windfall. Recent expansion and intensification of fragmentation, however, has occurred through anthropogenically-induced land use change. Habitat fragmentation is one of the most frequently cited threats to species and genetic diversity (Andren 1994), making it perhaps the most critical contemporary issue in conservation biology and one of the most important processes to understand if managed landscapes are to maintain their biodiversity (Wiens

1996, D'Eon et al. 2002). Effects of habitat fragmentation on populations and communities is a multifaceted process, occurring through a combination of habitat loss, reduction in habitat fragment sizes, and increasing isolation between fragments (Andren 1994, Forman 1995).

When fragmentation occurs, some portion of the original habitat is replaced by another land cover type, which is typically referred to as the landscape matrix. The matrix is often heterogeneous and may come to dominate the landscape (Forman 1995), where it can control species movements and natal dispersal (i.e., the dispersal of juveniles from the natal area to the site of first potential breeding; Greenwood 1980). The remnant habitat consists of a mosaic of isolated fragments (or patches) variously surrounded by, and perhaps interdigitated with, the surrounding matrix. Habitat loss itself has effects on species distributions, abundances, and richness; however, its effects are often exacerbated in fragmented landscapes by increasing inter-patch isolation and increased proportions of edge habitats

(Soule 1980, Andren 1994, Andren 1995, Bayne and Hobson 1998). In small mammal studies, habitat fragmentation has been linked to restricted movement and colonisation, 9 reduced population sizes and foraging abilities, increased predation and parasitism, and increased inbreeding depression and genetic drift (Levins 1970, Andren 1994, Forman 1995).

Fragmentation affects nearly all ecological patterns and processes (Forman 1995), but the effects are not consistently negative, especially for species that can use matrix habitats or thrive in edge habitats. For instance, highly vagile organisms with broad niches, such as raccoons (Procyon lotor L.), have been observed to derive some benefit from moderate amounts of forest fragmentation generated by agriculture (Pedlar et al. 1997). Despite growing concern over the potential impacts of habitat fragmentation, many of the underlying biological mechanisms and processes remain poorly understood. Forest fragmentation may have especially pronounced effects on arboreal forest specialists, such as the northern flying squirrel (Glaucomys sabrinus) and the red squirrel (Tamiasciurus hudsonicus); however, their populations have rarely been examined in agriculturally fragmented landscapes.

Glaucomys sabrinus are thought to play crucial roles in disseminating spores of mycorrhizal fungi that form important symbiotic relationships with many tree species (Maser

1979); they also serve as an important prey species for many predators, including fisher

(Martes pennanti Erxleben) and northern barred owls (Strix varia Barton) (Bayne and

Hobson 1998). Primarily due to its importance as prey for the endangered northern spotted owl (Strix occidentalis Merriam), G. sabrinus has been termed a keystone species in Pacific

Northwest forest ecosystems (Carey 2000). Northern flying squirrels are commonly considered to be habitat specialists, often relying on features abundant in mature coniferous forest habitats such as cavity trees, snags, downed woody debris, and multilevel forest canopies with high tree crown connectivity (Carey 1995, Carey et al. 1999, Holloway and

Malcolm 2006). Densities of G. sabrinus in central Ontario are strongly correlated with the 10 density of large spruce (Picea sp.) and hardwood trees, low levels of logging in the landscape, and snag density in conifer stands (Holloway and Malcolm 2006). Despite these findings, G. sabrinus may be habitat generalists throughout most of their range (Wheatley et al. 2005, Weigl 2007).

Tamiasciurus hudsonicus are widely considered to possess strong affinities and coevolutionary ties with coniferous forests throughout Canada (Goheen and Swihart 2005), yet they are found in a wide variety of forest types throughout their range (Swihart et al.

2007). Mature stands of coniferous forest are considered to be optimal habitat for T. hudsonicus and they attain their highest densities, greatest body mass, lowest mortality rates, and highest reproductive rates in these forests (Kemp and Keith 1970, Rusch and Reeder

1978, Obbard 1987). Tamiasciurus hudsonicus most commonly feed on the seeds of white spruce {Picea glauca), black spruce {Picea mariana), and jack pine {Pinus banksiana

Lamb.). As a result of their larder and scatter hoarding behaviours, T. hudsonicus plays a key ecological role in forest ecosystems by dispersing conifer seeds. Like G. sabrinus, T. hudsonicus is thought to have specific habitat requirements in Ontario. Holloway and

Malcolm (2006) found a strong relationship between T. hudsonicus density and characteristics of older, undisturbed coniferous forests, such as plentiful snags. Despite an apparent preference for conifer structure and features of mature forests, T. hudsonicus has managed to successfully expand its range into regions of the midwestern United States characterised by highly fragmented, second-growth deciduous forest (Goheen et al. 2003a,

Swihart et al. 2007).

Bruce and Grey counties, Ontario, have a mosaic of land uses, where towns, extensive road networks, and agriculture often create heavily fragmented forest habitats. As such, this 11 region lends itself well to mensurative, landscape-level studies of forest fragmentation and may be a prime candidate for appropriate management that ameliorates landscape effects.

Bruce and Grey counties, Ontario are in many senses ideal for fragmentation studies, containing a mix of fragmented and unfragmented forests along an approximately 200 km north-south swath. The forested landscape comprises approximately 22% of the total land area and forest patches vary widely in size and degree of isolation from neighbouring patches. The extensive reduction in forest area, combined with reduced patch size and increased isolation, presumably have caused changes in the local biota beyond what might be expected from habitat loss alone (Andren 1994). As a result, the spatial composition and configuration of habitat patches is likely to be an important predictor of species distributions in the region.

The effects of habitat fragmentation on wildlife species have often been studied within the general frameworks of either island biogeography theory (IBT; MacArthur and Wilson

1967) or the metapopulation concept (Levins 1969, Levins 1970). The metapopulation concept and IBT are both governed by the same ecological processes: extinction from and colonisation of patches (or islands, in the IBT context) of suitable habitat (Andren 1994).

Under both models, species diversity and population occurrence in a patch is a function of area and distance from a population source (i.e., isolation). No predictions are made with respect to species abundances under either theory; instead the two theories are best suited to predict species occurrence and species richness. Given this, occupancy data should serve as a useful surrogate of a population's viability and tolerance to fragmentation (Hanski 1994,

Laurance 1995, Vos et al. 2001, Swihart et al. 2006). Populations are considered to exhibit metapopulation structure if occupation of a patch is related to its size, while the probability of 12 patch recolonisation, following a local extinction event, is a function of the distance to occupied patches (Andren 1994). All else being equal, and given that the matrix is indeed unsuitable for both species, the expectation from IBT and the metapopulation concept is that

G. sabrinus and T. hudsonicus are more likely to occur in larger, less isolated forest fragments than in smaller, highly isolated patches.

Most often, studies of G. sabrinus and T. hudsonicus response to habitat fragmentation have focussed on patch area relationships in regions with >30% forest cover or regions not fragmented by impermeable matrix. Furthermore, studies of G. sabrinus and T. hudsonicus responses to habitat fragmentation have produced contrasting results across varying study contexts. For instance, Bayne and Hobson (1998) found a negative density-area relationship for G. sabrinus in 70-110 year old mixedwood boreal forest stands in Saskatchewan, Canada dominated by white spruce (Picea glauca) and trembling aspen (Populus tremuloides). Smith and Person (2007), however, reported a positive relationship between patch area and population persistence in old-growth (dominant trees >300 years old) forest stands composed predominately of sitka spruce {Picea sitchensis [Bong.] Carr.) and western hemlock (Tsuga heterophylla [Raf.] Sarg.) in Alaska. Moreover, southern flying squirrel (Glaucomys volans

L.) and Siberian flying squirrel (Pteromys volans L.) occurrence have been positively associated with increased patch area in Indiana and Finland, respectively (Nupp and Swihart

2000, Selonen and Hanski 2003). Likewise, T. hudsonicus density has been linked with reduced patch area in the mixedwood boreal forests of Saskatchewan, Canada (Bayne and

Hobson 2000), while T. hudsonicus occurrence has been linked with increased patch area in deciduous-dominated landscapes of the midwestern United States (Nupp and Swihart 2000,

Goheen et al. 2003b). There appear to be strong negative patch area-density relationships 13 across all tree squirrel species (Koprowski 2005), however patch area-occurrence relationships typically show up as being positive in nature (Nupp and Swihart 2000, Goheen et al. 2003&). While the reasons for these relationships are not completely clear, it is likely that tree squirrel densities are greater in smaller patches due to home range size compaction

(Koprowski 2005) and tree squirrel occurrence in large patches may be closely linked with increased population persistence (Selonen and Hanski 2003, Smith and Person 2007).

I am not aware of any studies relating G. sabrinus occurrence or abundance to parameters of interpatch isolation, but such studies that have been conducted on biologically similar species have not shown any significant relationship. For instance, in a study of G. volans in Indiana, Nupp and Swihart (2000) could find little evidence linking G. volans occurrence with multiple measures of interpatch isolation, including mean proximity index and nearest neighbour distance. Additionally, P. volans have been shown to disperse up to

8000 m in a fragmented boreal landscape in Finland using wooded corridors and deforested matrix (Selonen and Hanski 2004, Selonen and Hanski 2006), suggesting that interpatch distances may not limit patch occupancy in this species. Despite the aforementioned findings for G. volans and P. volans, G. sabrinus is widely considered to possess low vagility due to its gliding-based locomotion (D'Eon et al. 2002). Hence, the prevailing hypothesis is that G. sabrinus should be highly sensitive to interpatch isolation and low levels of landscape connectivity. Studies undertaken on the effect of interpatch isolation on T. hudsonicus occurrence and/or population abundance have also been unable to detect significant relationships (Bayne and Hobson 2000, Nupp and Swihart 2000, Goheen et al. 2003&).

Tamiasciurus hudsonicus is considered to possess good vagility, a factor which has been invoked to explain their successful range expansion into deciduous-dominated regions of the 14 midwestern United States in conjunction with intensified forest fragmentation (Goheen and

Swihart 2005). Investigations of G. sabrinus and T. hudsonicus dispersal in fragmented landscapes and their use of corridors and matrix have not yet been conducted.

Ecological attributes, such as the presence or absence of a species, depend entirely on both the habitat characteristics of focal patches and the characteristics of the landscape surrounding the focal patch (i.e., patch context or landscape structure) (Holland et al. 2004).

Multispecies, landscape-level studies are often conducted at a single spatial scale for all species under investigation, yet individual species likely respond differentially to the various aspects of their environment. More often than not, little is known a priori about the scale at which a species responds to patch context. Several researchers have indicated that the scales of response may be related to or approximated by the movement ranges of the organism under investigation (Addicott et al. 1987, Wiens and Milne 1989, Wiens et al. 1993, Cale and

Hobbs 1994, Vos et al. 2001, Dungan et al. 2002, Holland et al. 2005), but movement information is often lacking for most species (especially rare or poorly understood species).

Additionally, species may display divergent responses to different environmental variables at differing scales, hence multiple landscape variables should be investigated separately and the appropriate scale of study must be given proper consideration for each individual variable.

Despite their presumed ecological importance, very little is known about the sensitivity of G. sabrinus and T. hudsonicus to habitat fragmentation. The continued loss of forest in southern Ontario through fragmentation makes the conservation of G. sabrinus and T. hudsonicus metapopulations an important ecological issue. The focus of this study was to relate G. sabrinus and T. hudsonicus occurrence to parameters of habitat fragmentation considered to be fundamental to IBT and metapopulation dynamics, namely fragment size, 15 patch context (i.e., isolation) and habitat quality (i.e., the suitability of a fragments's habitat).

My objectives were to: (i) compare the occurrence of G. sabrinus and T. hudsonicus among a set of fragments to ascertain potential differences in habitat quality between sites and to search for habitat relationships for both species; (ii) to relate the occurrence of both species to various, biologically-relevant components of landscape structure at their appropriate scales; (Hi) determine appropriate landscape-level scales of investigation for both species as a function of their unique responses to patch context; and (iv) to establish minimum patch size requirements for both species, if area effects are shown to be present. I predict, based on IBT and the metapopulation concept, that the presence of both G. sabrinus and T. hudsonicus should correlate with increased patch size and reduced isolation. However, the habitat affinities apparent in both species may act as contributory factors affecting their occupancy. I suspect that the poor ground-based locomotion of G. sabrinus will act to impede their movement through agricultural matrix, and thereby increase their sensitivity to habitat fragmentation and, more specifically, patch isolation. The greater vagility of T. hudsonicus should result in less pronounced effects of isolation.

2. METHODS

2.1 Site Selection

I chose 24 forest fragments for study based on their deciduous-conifer tree composition, fragment area, proximity to other fragments, and access (Figure 1.1). Thirty potential sites were initially identified using a digital, 28 m resolution Ontario Ecological 16

Land Classification (OELC) map provided by the Ontario Ministry of Natural Resources

(OMNR) and 1:10000 orthogonal aerial photographs provided by Grey-Sauble and Saugeen

Valley Conservation Authorities. Based on initial visits to the 30 potential sites and consultations with Grey-Sauble and Saugeen Valley Conservation Authority staff, I reduced the number of sites to 24. To the extent possible, I chose forests that, based on habitat relationships, provided suitable habitat for the species. All fragments were upland forest that was deciduous-dominated (mean % deciduous = 80% and range = 50 to 99%, from basal area measurements [see below]), as this was the pre-eminent forest type in this study area.

Fragments were chosen to span the range of patch sizes found in the study area with a predetermined minimum patch size of 3.84 ha needed to establish the trapping grid (mean patch size = 240 ha, range = 4 to 2881 ha) and spanned a range of isolations (mean nearest neighbour distance range = 187 m to 1968 m, see Methods section below for more detail). To improve spatial independence, the minimum distance between sites was approximately 10 km (mean inter-site Euclidean distance = 45.2 km and range = 9.7 to 98.9 km). Sites experiencing active or recent anthropogenic disturbances, such as timber harvest (as evidenced from stumps and consultation with conservation authority staff) and cattle grazing, were avoided.

2.2 Sciurid Trapping

Live-trapping was conducted between June 1 and September 1, 2006. Fragments were sampled by use of a 4-by-9 grid, with 40 m spacing between trap stations. A trap station consisted of a single large-sized Sherman trap (10-by-12-by-38 cm) attached to the nearest 17

^ Sites

sis

20 km

Figure 1.1: Map of Grey and Bruce counties, Ontario with sites where presence/absence of Glaucomys sabrinus and Tamiasciurus hudsonicus marked (stars). Two Universal Trans verse Mercator, Zone 17 grid lines also are shown (map datum: NAD83). 18 tree >20 cm diameter at breast height (DBH) and at a height of approximately 1.5 m. Traps were mounted on top of two 15 cm long spikes driven into the tree stem and secured with an elastic cord (see Holloway and Malcolm 2007). I pre-baited each trap station for three nights prior to the first trap-night by placing peanut butter on the tree trunk. Traps were baited with peanut butter, soaked sunflower seeds, and a slice of apple and set for three consecutive nights during each trapping session and reset each morning. Unbleached cotton was provided as bedding.

Each forest fragment was trapped during two trap sessions, with approximately six weeks between sessions (June 1-July 8 and July 9-September 1). Therefore, each forest fragment was trapped for a total of six nights (216 trap-nights). All captures were marked with Monel No. 1 metal eartags (National Band and Tag Co., Newport, KY) and weight, reproductive status, hind foot length, and tail length were measured. Trapping and animal handling procedures were approved by the Animal Care Committee of the University of

Toronto.

2.3 Habitat Measurements

Previous work has indicated that important habitat features for G. sabrinus and T. hudsonicus include deciduous-coniferous forest composition (Vahle and Patton 1983,

Holloway and Malcolm 2006, Holloway and Malcolm 2007), snag abundance (Weigl 1978,

Fancy 1980, Vahle and Patton 1983, Carey et al. 1997, Holloway and Malcolm 2007), canopy openness (Sullivan and Moses 1986, Carey 2002, Holloway 2006), subcanopy forest structure (Hackett and Pagels 2003), and downed woody debris abundance (DWD) (Carey et 19 al. 1997, Hackett and Pagels 2003), hence my measurements focused on these habitat features. I used prism sweeps (BAF2) to quantify forest compositional and structural characteristics at each of the 24 sites. Prism sweeps were made at alternating trap stations for a total of 18 sweeps per site. I measured all trees and snags with a DBH greater than 10 cm, classified them as canopy or subcanopy, and identified them to the species level (where possible for snags). Subcanopy trees were classified as trees with crowns entirely below the general level of the canopy. The basal areas (m2/ha) of all living conifer and hardwood trees, canopy and subcanopy trees, and snags, were calculated for each site. I measured the percentage of canopy closure with a concave spherical densiometer, averaging 4 readings in each of the cardinal directions at alternating trap stations for a total of 18 readings per site.

Downed woody debris was measured within a 10 m radius plot centred on the trap station tree. Plots were established at alternating trap stations for a total of 18 plots per site (and a total sampled area of 5655 m2 per site). The diameter of each piece of DWD (minimum diameter 10 cm) contained within a plot was measured at the midpoint of the portion of the log within the plot (excluding portions of the log <10 cm diameter). Diameters were summed across the entire site. Each piece of DWD was classified into one of two decay groups based on Maser et al. (1979): early decay (decay classes 1-3) or advanced decay

(decay classes 4 and 5). To improve normality and reduce heteroscedasticity in the habitat data, summed diameters of early and advanced decay class groupings were logio transformed.

2.4 Landscape Measurements

Landscape structure was measured by use of a digital, 1:5000 scale map indicating 20 forest polygons with a positional accuracy of +/-10 m (OMNR, unpublished). This vector map had a minimum mapping unit of 0.25 ha where mapped from orthogonal aerial photography and a minimum of 0.5 ha where mapped from satellite imagery. All wooded areas were mapped that met the following standards: trees > 2 m in height and 60% canopy coverage (OMNR, unpublished). Attributes of the vector layer were limited to forested and, by default, non-forested (i.e., matrix) land cover.

The vector map was rasterised in ArcMap 9.2 (Environmental Systems Research

Institute 2007) to a pixel size of 28 m and to the same pixel boundaries (and projection

[NAD83, Lambert Conformal Conical]) as the OELC raster map. Using the 8-neighbour rule, all forest cells adjacent to a given forest cell were considered to be members of the same patch. The rasterised forest map was then overlaid onto the OELC raster map to include all water pixels. Overlaid map layers were visually inspected to ensure a proper alignment of features.

In the absence of a single, widely accepted metric that adequately captures all aspects of habitat fragmentation, it is generally considered best practice to measure fragmentation by use of several indices in tandem (Hargis et al. 1998). In addition to logio-transformed patch area (logl0_area), I chose to measure the following fragmentation indices within circular landscape windows centred on the trapping grid: mean proximity index, mean nearest neighbour distance (meannnd), number of patches (npatch), and proportion of forest

(pr_forest) (see following sections for details). These metrics were selected because they expressed important aspects of landscape structure, including the amount of forest in the surrounding area (proportion of forest, number of patches) and its spatial configuration (mean proximity index, mean nearest neighbour distance) (Verboom and van Apeldoorn 1990, 21

Fitzgibbon 1993, van Apeldoorn et al. 1994, Wauters et al. 1994, Bayne and Hobson 1998,

Nupp and Swihart 1998, Nupp and Swihart 2000, Selonen et al. 2001, Vos et al. 2001,

Swihart et al. 2006). The selected metrics are often cited as ecologically important fragmentation indices (Gustafson and Parker 1992, Andren 1994, Paton 1994, McGarigal and

Marks 1995, Bender et al. 1998, Davidson 1998, Vos et al. 2001). All the landscape metrics were derived from the overlaid raster map described above, except for patch area, which was obtained from the original vector map. Circular landscape windows had radii that ranged from 1000 to 10000 m, at 1000 m intervals (see below for a justification of this range).

Metrics were calculated by use of a SAS program that processed an ASCII version of the raster map (Malcolm and Patterson, unpublished).

2.4.1 Mean Proximity Index

Mean proximity index quantifies the degree of isolation of a focal habitat fragment in relation to its neighbours. According to Gustafson and Parker (1994), it is most useful in distinguishing sparse distributions of small habitat fragments or isolated fragments from configurations of clustered, large fragments. It is considered an effective measure of patch isolation for remnant or disturbed patches (Hargis et al. 1998) and is assumed to be a good indicator of colonisation dynamics as predicted by the metapopulation concept and IBT

(Gustafson and Parker 1994). The mean proximity index is calculated using the area (S) and edge-to-edge nearest neighbour distance (z) of the focal fragment (/) to each of the i fragments within the specified buffer radius (Whitcomb et al. 1981) and is measured from cell centre to cell centre: 22

Mean Proximity Index = 1^

The approach used here differs from a similar metric developed by Gustafson and

Parker (1992) in that I used the nearest neighbour distances between edges of the focal patch and all neighbouring patches in the circular landscape window {sensu Whitcomb et al. 1981,

McGarigal and Marks 1995) instead of the nearest neighbour distances between all patches within the circular landscape window. The mean proximity index used here is dimensionless and can be used as a comparative index as long as the search buffer radius used around each patch is identical (Gustafson and Parker 1992).

2.4.2 Mean Nearest Neighbour Distance

Mean nearest neighbour distance is one of the most widely used measures of patch isolation in landscape ecology studies because of its simplicity (McGarigal and Marks 1995).

Within a specified search radius, it measures the average edge-to-edge Euclidean distance (z) between a focal fragment (/) and its neighbouring fragments (i), where n is the total number of fragments in the circular landscape window:

Mean Nearest Neighbour Distance = —— n 23

2.4.3 Proportion of Forest

Proportion of forest measures the total proportion of suitable habitat within a specified search radius. Because the rasterised landscape map contained only three classes (forest, water and matrix), the proportion of forest within the specified search radii was defined as:

Proportion of Forest = Circular Landscape Window Area - Water Area J

2.4.4 Number of Patches

Number of patches was computed as the total number of non-focal patches within the specified search radius.

2.5 Statistical Analysis

2.5.1 Landscape Scale

One approach to estimating appropriate scales of analysis is to model the relationship between a response variable and a predictor variable at several scales and determine the scale that results in the greatest model fit (Holland et al. 2004). To do this, for each species, a plot of Akaike Information Criterion (AICc) model fit statistics against spatial scale (landscape window radius) was produced for each variable of interest. The scale corresponding to the best logistic model fit (lowest AICc value) was selected as the most appropriate scale for 24 investigation.

In the absence of a priori information on how G. sabrinus and their populations respond to spatial scales, I used ten landscape radii (1000 to 10000 m at 1000 m intervals).

The Siberian flying squirrel (P. volans) has been shown to disperse up to a maximum distance of c. 9000 m in contiguous forest and c. 8000 m in fragmented forests (mean dispersal distance = 2500 m) during natal dispersal (Selonen and Hanski 2004, Selonen and Hanski

2006). Given that body mass and dispersal distance are related in mammals (Wolff 1999,

Sutherland et al. 2000) and that G. sabrinus and P. volans share similar body mass and are both gliders, it is possible that G. sabrinus may be capable of similarly dispersing over large distances in fragmented landscapes. Dispersal distance also has been shown to be proportional to home range size in mammals and home range size may serve as a better predictor of dispersal distance than body size (Bowman et al. 2002). Following Bowman et al.'s (2002) finding that dispersal distance can be predicted as a simple multiple of the linear dimension of home range size (i.e., the square root of home range area), I calculated the predicted maximum and median dispersal distances for G. sabrinus given a mean home range size of 10.2 ha (from Holloway [2005] in central Ontario). The square root of this home range area is 319 m, which, following Bowman et al. (2002), yields an estimated maximum dispersal distance of 12776 m and a median dispersal distance of 2236 m. Accordingly, I used a range of 1000-10000 m, which approximately spans these estimates.

Unlike G. sabrinus, there is considerable information on the natal and post-breeding dispersal distances of T. hudsonicus. Generally, dispersal distances are relatively short in T. hudsonicus populations, with a maximum observed natal dispersal distance of 4500 m

(Haughland and Larsen 2004a, Haughland and Larsen 2004&). Long-range natal dispersion 25 has also been documented at 2300 m (Haughland and Larsen 2004a), 1090 m (Sun 1997) and

922 m (Larsen and Boutin 1994). Establishment of natal territories usually occurs within 150 m of natal sites, although exploratory movements of over a kilometre appear to be common

(Larsen and Boutin 1995). Post-breeding dispersal distance of female T. hudsonicus who bequeath their territories to their offspring averages 82 m (maximum 144 m). Despite these indications of smaller dispersal distances for T. hudsonicus than G. sabrinus, I used the same set of landscape radii to permit comparison between the two species.

The inclusion of all search radii in the analysis presented a practical problem: at the larger radii, landscapes in some cases overlapped with each other, potentially compromising the statistical independence of the data. To investigate this possible problem, I explored scale relationships in two ways. In one, I carried out logistic regressions on species occurrence for each landscape predictor (proportion of forest, number of patches, mean nearest neighbour distance, mean proximity index) at the various search radii by use of the computer program

Focus (Holland et al. 2004), which was set to randomly select 100 sets of spatially independent sites for each search radius (i.e., sites whose landscape windows did not overlap). In another, I used all 24 sites for the analyses irrespective of landscape overlap. For each analysis, I plotted mean AICc values against the landscape radius for each variable. In addition, to examine pattern similarity between the two approaches, I calculated Spearman correlation coefficients between the two sets of AICc values ordered according to landscape window radius. These two approaches yielded very similar patterns (see results); hence in further analyses I undertook bivariate and multivariate approaches that used all 24 sites.

2.5.2 Bivariate and multivariate approaches 26

In addition to the logistic regressions described above, I also undertook logistic regressions for each of the eight habitat variables and five landscape structure variables.

Based on these regressions, I took the two most significant landscape structure variables and the two most significant local habitat variables, and investigated their joint contributions through multiple logistic regression. Small sample size (n=14 fragments with G. sabrinus present and n=12 fragments with T. hudsonicus present) did not permit me to simultaneously include all variables. I employed forward selection with significance for entry set to 0.05 and significance for removal set to 0.1.1 asked whether habitat variables had any significant information given that landscape variables were forced into the model; similarly, I examined the significance of the landscape variables given that the habitat variables were forced into the model. Predictive models were then produced for G. sabrinus and T. hudsonicus occurrence using logistic regression with a forward selection procedure (significance for entry set to 0.05 and significance for removal set to 0.1). Finally, I developed a receiver operating characteristic (ROC) curve for G. sabrinus based on the probability of patch occupancy since that patch area was the proximate factor controlling occupancy. The purpose of the ROC curve was to identify the threshold patch area requirement for G. sabrinus that maximised the modelled true-positives while minimising the modelled true-negatives.

In order to explore and identify correlations between the independent landscape structure variables I undertook principal component analysis (PCA) on the suite of variables.

This ensured that correlated variables were not jointly entered into bivariate and multivariate procedures. I used a reduced set of uncorrelated landscape and habitat variables to investigate occurrence of each of the two species via redundancy analysis (RDA). In this analysis, the vector of species occurrence was used as the "species matrix" and the habitat and/or landscape variables used as the "environment matrix". I tested for an effect of the habitat variables with the landscape variables partialled out (i.e., controlled for) as covariates (i.e., tested for an effect of habitat given that the landscape structure had already been accounted for), and vice versa, for both species. These results were then combined with an RDA that included all uncorrelated habitat and landscape variables such that the unexplained variance explained could be partialled out in a variance partitioning procedure. RDAs were carried out with 9999 Monte Carlo unrestricted permutation tests and species scores were centred and standardised.

Finally, because of the importance of patch area for G. sabrinus, I tested the predictive significance of the landscape variables individually with patch area partialled out in an RDA.

Variance explained and p-values were compared with those from similar RDAs in which patch area was not partialled out. This served to test for a predictive effect of the landscape structure variables independent of patch area effects. RDAs were carried out with 9999

Monte Carlo unrestricted permutation tests and species scores were centred and standardised.

Logistic regressions and the ROC curve were carried out using SAS software, logistic regression response curves were produced in R (R Development Core Team 2007), and

CANOCO (Ter Braak and Smilauer 1998) was used to undertake PCAs and RDAs. In all cases, I used AICc values, which is a correction of standard AIC values for small sample sizes (Burnham and Anderson 2002).

3. RESULTS 28

3.1 Sciurid Occurrence

In total, I captured 48 G. sabrinus (0.93/100 trap nights) and 96 T. hudsonicus

(1.85/100 trap nights) individuals during the 5184 trap nights. Glaucomys sabrinus was present at 14 and T. hudsonicus at 12 of the 24 sites.

3.2 Spatial Independence

The AICc values from the spatially independent (18 sites) and spatially non- independent (24 sites) were all highly correlated and showed the same minima (i.e., best model fit) in all cases. For G. sabrinus, Spearman correlation coefficients (rs) for the four variables (mean proximity index, mean nearest neighbour distance, proportion of forest, and number of patches) were, respectively, 0.71 (p=0.0221), 0.97 (p=0.0001), 0.78 (p=0.007), and 0.85 (p=0.0017) (Figure 1.2). Similarly, respective Spearman correlation coefficients (rs) for T. hudsonicus were 0.98 (p=0.0001), 0.88 (p=0.0007), 0.88 (p=0.0009), and 0.92

(p=0.0001) (Figure 1.3). These results suggest that the scales were not spatially dependent enough to affect the patterns of explanatory power as a function of landscape radius.

Accordingly, I included all 24 sites in further analyses to make use of all available data and maximise statistical power.

3.3 Landscape Scale

The landscape radii that showed the best model fit for G. sabrinus were 7000 m for Mean proximity index Mean nearest neighbour distance Proportion of forest Numer of patches

sir 38

29 T 32 36

34 w*TT*^ h^J ^" p = 0.007, r =0.78 32 p = 0.0017, r =0.85 p = 0.000 l,r, = 0.97 a 30 si y y \ * < < 28 . v^. /N • 26 v N/ * 24

i 22 201 20 Radius (m) Radius (m) Radius (m) Radius (m) 124 Sites «18 Sites Figure 1.2: Corrected Akaike Information Criterion (AICc) values from logistic regressions between Glaucomys sabrinus occurrence and mean proximity index, mean nearest neighbour distance, proportion of forest, and number of patches as a function of landscape radius for spatially independent (18 sites, lower line plot) and non-spatially inde pendent (24 sites, upper line plot) data sets. The radius of best model fit is indicated (drop-down arrow) and Spear man correlation results (p and r values) between the spatially independent and non-spatially independent AICc values are given.

to Mean proximity index Mean nearest neighbour distance Proportion of Forest Number of patches

MM I ^^*-M^*^ 36 (l»-*-- 1 ( 341i ] p = 0.0001, r=0.98 p = 0.0001, rs = 0.92 32 it p = 0.0007, r=0.i j H a 30i < »-• • • «^ -

•—# 0 m rY •~i j

f f • j

<* s*> « Radius (m) Radius (m) Radius (m) Radius (m) 124 Sites «18 Sites

Figure 1.3: Corrected Akaike Information Criterion (AICc) values from logistic regressions between Tamiasciurus hudsonicus occurrence and mean proximity index, mean nearest neighbour distance, proportion of forest, and number of patches as a function of landscape radius for spatially independent (18 sites, lower line plot) and non- spatially independent (24 sites, upper line plot) data sets. The radius of best model fit is indicated (drop-down arrow) and Spearman correlation results (p and r values) between the spatially independent and non-spatially in dependent AICc values are given. 31 mean proximity index (AICc=29.5), 7000 m for mean nearest neighbour distance

(AICc=26.3), 3000 m for proportion of forest (AICc=28.6), and 9000 m for number of patches (AICc=35.0) (Figure 1.2). Respective radii for T. hudsonicus in all cases were considerably smaller; specifically, they were: 2000 m (AICc=37.0), 2000 m (AICc=35.6),

3000 m (AICc=37.0), and 3000 m (AICc=36.7) (Figure 1.3).

3.4 Logistic Regressions

Of the landscape variables, mean nearest neighbour distance had a significant negative relationship with G. sabrinus occurrence (p=0.028), and proportion of forest (p=0.025) and area (p=0.016) both had significant positive relationships (Table 1.1). Mean proximity index approached significance (p=0.061), but number of patches was not a significant variable

(p=0.252). None of the habitat variables were significant predictors of G. sabrinus occurrence (all /?>0.133, Table 1.1).

In contrast, none of the landscape variables were significant predictors of T. hudsonicus occurrence (Table 1.2). Of the local habitat variables, basal area of deciduous trees was a significant negative predictor (p=0.046) and basal area of coniferous trees a significant positive predictor (p=0.047); none of the other habitat variables were significant (all /?>0.213;

Table 1.2).

Logistic regression models with a forward selection procedure did not reveal any significant effect of the best two uncorrelated habitat variables (advanced decay class downed woody debris [p=0.13] and basal area of deciduous trees [p=0.21]) in predicting G. sabrinus occurrence when patch area (p=0.03) and proportion of forest (/?=0.14) were already in the 32

Table 1.1: Attributes for landscape and habitat variables in Bruce and Grey counties, Ontario, for Glaucomys sabrinus. Slope, Pr(>lzl) and corrected Akaike Information Criterion values are from logistic regression analysis for each individual predictor variable. Scales of best model fit and transformations are given in brackets beside corresponding terms. BOLD indicates significant values (p < 0.05). Mean values (± 1 SD) are given separately for sites where Glaucomys sabrinus was present and absent.

Attributes Mean±SD Mean±SD Slope Pr(>lzl) AICc (Presence) (Absence) (n = 14) (II = 10)

Landscape Structure Mean Proximity Index (7000 m) 1824.22 ± 484.98 381.85 ± 164.25 1.87 0.061 29.5 Mean Nearest Neighbour Distance (m) (7000 m) 3068.33 ± 1207.42 4177.09 ± 407.55 -2.20 0.028 26.3 Proportion of Forest (3000 m) 0.48 + 0.19 0.27 + 0.14 2.25 0.025 28.6 Number of Patches (9000 m) 669 ±470 475 ±267 1.15 0.252 35.0 Area (ha) (logio) 2.19 ±0.53 1.37 + 0.43 2.42 0.016 21.6

Habitat (Stand Level) % Canopy Closure 78 ±6 80 + 8 -0.79 0.431 36.1 Basal Area of Canopy Trees (m2/ha) 21.2 ±4.0 22.1 ±2.5 -0.63 0.526 36.4 Basal Area of Subcanopy Trees (m2/ha) 6.9 ±3.1 6.3 ±4.4 0.40 0.687 36.6 Basal Area of Deciduous Trees (m2/ha) 20.5 ±5.1 23.4 ±3.2 -1.50 0.133 34.1 Basal Area of Coniferous Trees (m2/ha) 6.7 ±5.3 4.1 ±3.5 1.34 0.180 34.8 Basal Area of Snags (m2/ha) 2.4 ±0.9 2.3 ±0.9 0.34 0.733 36.7 Summed Diameter of Early Decay Class 3.9 ±0.2 3.9 ±0.3 -0.26 0.795 36.7 Downed Woody Debris (cm) (logio) Summed Diameter Advanced Decay Class 4.3 ±0.4 4.0 ±0.4 1.44 0.151 34.5 Downed Woody Debris (cm) (logio) 33

Table 1.2: Attributes for landscape and habitat variables in Bruce and Grey counties, Ontario, for Tamiasciurus hudsonicus. Slope, Pr(>lzl) and corrected Akaike Information Criterion values are from logistic regression analysis for each individual predictor variable. Scales of best model fit and transformations are given in brackets beside corresponding terms. BOLD indicates significant values (p < 0.05). Mean values (± 1 SD) are given separately for sites where Tamiasciurus hudsonicus was present and absent.

Attributes Mean ± SD Mean ± SD Slope Pr(>lzl) AICc (Presence) (Absence) (n = 12) (n=12)

Landscape Structure Mean Proximity Index (2000 m) 145.49 ± 57.87 183.90 ± 58.57 -0.48 0.629 37.0 Mean Nearest Neighbour Distance (m) (2000 m) 664.99 ± 322.01 836.14 ±338.60 -1.24 0.213 35.6 Proportion of Forest (3000 m) 0.33 ±0.16 0.295 ±0.130 0.55 0.581 37.0 Number of Patches (3000 m) 94 ±72 113 + 62 -0.72 0.470 36.7 Area (ha) (logio) 2.01 ± 0.66 1.691 ±0.60 1.21 0.227 35.8

Habitat (Stand Level) % Canopy Closure 77 ±7 80 ±7 -1.05 0.293 36.3 Basal Area of Canopy Trees (m2/ha) 21.1+4.1 22.1 ±2.7 -0.68 0.494 37.0 Basal Area of Subcanopy Trees (m2/ha) 7.02 ± 3.9 6.2 ± 3.5 0.56 0.578 37.1 Basal Area of Deciduous Trees (m2/ha) 19.7 ±4.6 23.7 ± 3.6 -1.99 0.047 32.1 Basal Area of Coniferous Trees (m2/ha) 7.7 ± 4.9 3.6 ±3.9 1.99 0.046 32.4 Basal Area of Snags (m2/ha) 2.6 ± 0.7 2.2 ±1.1 1.02 0.310 36.4 Summed Diameter Early Decay Class Downed 3.9 ±0.2 3.9 ±0.3 0.58 0.562 37.1 Woody Debris (cm) (logio) Summed Diameter Advanced Decay Class Downed 4.2 ± 0.5 4.2 ± 0.4 -0.04 0.970 37.5 Woody Debris (cm) (logio) 34 model (Table 1.3). Similarly, no significant effect of the best two landscape variables (patch area [p=0.135] and mean nearest neighbour distance [p=0.173]) were detected with respect to

T. hudsonicus occurrence when the inclusion of both basal area of deciduous trees (p=0.271) and percent canopy closure (p=0.345) were forced into the model (Table 1.4).

Significant logistic regression models were developed using the two most significant landscape and the two most significant habitat variables in multiple logistic regressions with a forward selection procedure to determine the most proximate factor predicting the presence of G. sabrinus and T. hudsonicus. Probability of occurrence of G. sabrinus exhibited a significant positive relationship with patch area (p=0.016, Figure 1.4), while T. hudsonicus showed a significant negative association with basal area of deciduous trees (p=0.047, Figure

1.5). The Hosmer-Lemeshow test (Hosmer and Lemeshow 2000) showed no evidence for a lack of model fit in the G. sabrinus and T. hudsonicus models (p=0.90 and p=0.69, respectively). Further, 93% and 74% of the actual and modelled pairs were found to be concordant in the G. sabrinus and T. hudsonicus models, respectively, and the likelihood ratio test (p=0.0001 and/?=0.021, for G. sabrinus and T. hudsoncius, respectively) indicated that at least one of the predictors' regression coefficients was not equal to zero in either model.

The modelled probability of G. sabrinus patch occupancy was plotted as a receiver operating characteristic (ROC) curve to determine the threshold patch area requirement for

G. sabrinus occupancy (Figure 1.6). The area under the ROC curve was given by the c statistic, which was 0.932, indicating a good model fit (a c value of 1 is maximum model fit).

The ROC curve shows that the threshold cut-off point (i.e., where sensitivity is maximised and 1 - specificity is minimised [at the point closest to the upper left corner of the plot]) is at the probability threshold of 0.46. This probability returns a threshold patch area requirement 35

Table 1.3: Effect of the most significant habitat variables on predicting Glaucomys sabrinus incidence in Grey and Bruce counties, Ontario with the inclusion of the most significant landscape variables forced into a multiple logistic regression model. Significance level for entry into the model was a = 0.1; for exit it was 0.15. Characteristic scales of response and transformations are given in brackets beside corresponding terms.

Pr(>lzl) Analysis of Effects in the Model Proportion of Forest (3000 m) 0.140 Area (ha) (logio) 0.030

Analysis of Effects Not in the Model Summed Diameter Advanced Decay Class Downed Woody Debris (cm) 0.130 (logio) Basal Area of Deciduous Trees (m2/ha) 0.209

Testing Global Null Hypothesis: BETA=0 Likelihood Ratio Test 0.0001 Score Test 0.006 WaldTest 0.070

Hosmer and Lemeshow Goodness-of-Fit Test 0.955 36

Table 1.4: Results of redundancy analyses evaluating the significance of each landscape variable on predicting Glaucomys sabrinus occurrence without and with patch area partialled out (i.e., area entered as a covariate). Monte Carlo statistical tests were run with 9999 unrestricted permutations and species scores were centred and standardised.

Pr(>lzl) Analysis of Effects in the Model Basal Area of Deciduous Trees (m2/ha) 0.271 Basal Area of Coniferous Trees (m2/ha) 0.345

Analysis of Effects Not in the Model Area (ha) (logio) 0.135 Mean Nearest Neighbour Distance (2000 m) 0.173

Testing Global Null Hypothesis: BETA=0

Likelihood Ratio Test 0.043 Score Test 0.059 WaldTest 0.100

Hosmer and Lemeshow Goodness-of-Fit Test 0.062 V3 3

C/5 I O

4) O s o3 o o

Patch area (logl0)

Figure 1.4: Presence/absence of Glaucomys sabrinus in relation to the loglO-transformed patch area in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region, with a fitted probability of re sponse curve derived from the logistic regression model (p = 0.016, AICc = 21.5). 38

15 20 25 Basal area of deciduous trees (m2/ha)

Figure 1.5: Presence/absence of Tamiasciurus hudsonicus in relation to the basal area of deciduous trees in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region, with a fitted probability of response curve derived from the logistic regression model (p = 0.047, AICc = 31.94) i.ol

0.9

0.8 1

0.7

*-» 0.6 1

0.5 m C 0.4 1 00

0.3

0.2 1

0.1

0.0 -i 1—i r- "1 1 1 r- -T—i—j—r—r™ 0.0 O.t 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 - Specificity

Figure 1.6: Receiver operating characteristic (ROC) curve based on a logistic re gression model for the presence/absence of Glaucomys sabrinus as a function of the probability of Glaucomys sabrinus occurrence given log 10 patch area in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region. for G. sabrinus of 46.5 ha when entered into the logistic regression model.

3.5 Multivariate Approaches

Principal component analysis was used to determine correlations among the independent landscape structure variables at their respective scales of best model fit. For G. sabrinus occurrence, mean nearest neighbour distance and mean proximity index were negatively correlated (Figure 1.7). Similarly, for T. hudsonicus occurrence, mean nearest neighbour distance and mean proximity index were negatively correlated (Figure 1.8).

Additionally, for T. hudsonicus, number of patches and proportion of forest were positively correlated at their respective scales (Figure 1.8). None of the correlated variables were jointly entered into any of the bivariate or multivariate procedures.

When G. sabrinus occurrence was constrained by the predictor variable, the suite of habitat variables explained 4% of the variance (p=0.44), whereas the suite of landscape variables explained 22% of the variance (p=0.0012). When a variance partitioning procedure was conducted, the landscape variables and habitat variables combined to explain 44% of the total variation in the species data when both were entered as explanatory variables in an

RDA. Meanwhile, the suite of landscape variables explained 23% (£>=0.0019) of the variance when the habitat variables were entered as covariables, and the habitat variables explained

9% (p=0.075) of the variance when the landscape variables were entered as covariables.

Therefore, independently the landscape and habitat effects explained a combined 32% of the variation in the species data and, thus, the remaining 12% of the variation was shared between landscape and habitat. 41

p npatch : 15/ ; •/ 13

12 prjorest 14 ; • meannnd 5 4 : • • / 8 logl0_area : D 20 7 SO 23* * 1lfl« • 6 • C3 - • 22 18 D a #16 9 17 10 • proximity 24 D 19 2

D 1 p

1 h- —i 1.0 1.0 Absence Presence

O T5 D npatch / prjorest 12 •

//* 23 13 8 D meannnd • D

1 14 V 18 .. D •

• 10a 24* D 6 logJO_area D 20 16 • 9

proximity n • • 19 17 E 1

D 2 a 5 1 1— 1 -1.0 1.0 Absence Q] Presence Figure 1.8: Principal component analysis of all independent landscape variables at their respective scales of best model fit showing sites where Tamiasciurus hudsonicus was present and absent in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region. Inter-sample scaling was used and species scores were centred and standardised. 43

Tamiasciurus hudsonicus showed a contrasting pattern. The habitat variables explained

12% of the variance and were a significant predictor of T. hudsonicus occurrence (p=0.04); in contrast, the landscape variables explained only 4% of the variance and were not a significant predictor (/?=0.50). When a variance partitioning procedure was conducted, the landscape variables and habitat variables explained 29% of the total variation in the species data when both were entered as explanatory variables in an RDA. Meanwhile, the suite of landscape variables explained 4% (/?=0.43) of the variance when the habitat variables were entered as covariables, and the habitat variables explained 17% (p=0.035) of the variance when the landscape variables were entered as covariables. Therefore, independently landscape and habitat effects explained a combined 21% of the variation in the species data and, thus, the remaining 8% of the variation was shared between landscape and habitat.

Finally, for G. sabrinus occurrence, none of the landscape composition or landscape configuration variables were significant when area was partialled out (Table 1.5).

4. DISCUSSION

4.1 Scale of best model fit

In studies of species responses to environmental variables, it is important that an appropriate spatial scale be used. The strength of the relationship between species responses and the environmental variable of interest may vary across a range of different scales.

Further, species responses to their environment are likely species-specific and variable- 44

Table 1.5: Results of redundancy analyses evaluating the significance of each landscape variable on predicting Glaucomys sabrinus incidence without and with patch area partialled out (i.e., area entered as a covariate). Monte Carlo statistical tests were run with 9999 unrestricted permutations and species scores were centred and standardized.

Pr(>lzl) Variance Explained meannnd without loglO_area partialled out 0.004 27% proximity without loglO_area partialled 0.011 15% out pr_forest without loglO_area partialled out 0.018 19% npatch without loglO_area partialled out 0.150 9% meannnd with loglO_area partialled out 0.849 1% proximity with loglO_area partialled out 0.899 1% pr_forest with loglO_area partialled out 0.453 2% npatch with loglO_area partialled out 0.446 3% 45 dependent, so conducting multi-species and/or multi-variable landscape-level studies at a single scale should be avoided. In this study, I used one published method to determine the appropriate spatial scale for analysis based on the best logistic regression model fit between species occurrence and the environmental variable of interest across a range of spatial scales

(Holland et al. 2004).

Holland et al. (2004) stated that the spatial scale of best model fit may in fact be related to the movement range(s) (i.e., dispersal distance) of the organism under investigation.

Explicit studies of G. sabrinus movement and dispersal are lacking, but it is possible that the scales of best model fit found here (3000 m and 7000-9000 m) correspond roughly to mean and maximum dispersal distances of G. sabrinus. As discussed in Methods (section 2.5.1) above, the biologically similar Siberian flying squirrel (P. volans) has been observed to disperse a maximum distance of 8000 m during natal dispersion in a fragmented boreal forest landscape in Finland, with a mean dispersal distance of 2500 m (Selonen and Hanski 2004,

Selonen and Hanski 2006). These findings are not unlike the scales of best fit found here for

G. sabrinus. Likewise, the two scales of best model fit for T. hudsonicus (2000 m and 4000 m) correspond with the range of maximum natal dispersal distances observed by Haughland and Larsen (2004a, 20046): 2300 m - 4500 m. Thus, I find support for the idea that the most appropriate spatial scale for investigation is best approximated by, or at the very least related to, the movement range of the focal organism (Addicott et al. 1987, Wiens and Milne 1989,

Wiens et al. 1993, Cale and Hobbs 1994, Vos et al. 2001, Holland et al. 2004). Should the relationship between dispersal distance and scale of response be found to be true through empirical testing, then I have found evidence that G. sabrinus dispersal distances are greater than those of T. hudsonicus in fragmented secondary hardwood Great Lakes-St. Lawrence forests, as was predicted here a priori based on studies of P. volans dispersal and home range size-dispersal distance relationships. However, I must caution against using dispersal distance information to approximate scales of investigation until further studies are conducted to confirm this trend, especially in fragmented landscapes where dispersal may be limited by edge and matrix effects on animal behaviour.

It is also of interest that 71 hudsonicus did not show a strong response to any of the studied landscape variables as a function of landscape scale. This suggests that T. hudsonicus is not notably affected by changes in landscape composition or landscape configuration at any scale greater than 1000 m, and that the effects of forest fragmentation on T. hudsonicus occurrence should be minimal.

4.2 Responses ofG. sabrinus to habitat features

Northern flying squirrels were found in 60% (14 of 24) of the forest fragments sampled in this study. Presence of G. sabrinus was found to be statistically unrelated to site-level habitat features, including canopy closure, deciduous-coniferous composition, canopy- subcanopy composition, and availability of snags and downed woody debris. Three of the most frequently cited habitat resources for G. sabrinus are abundant snags, the presence of conifer structure, and stands of mature or old-growth forest (Carey 1989, Carey 1991, Witt

1992, Ransome and Sullivan 1997, Carey et al. 1999, Carey 2000, Smith et al. 2004,

Holloway and Malcolm 2006); however, I could find little evidence that G. sabrinus occurrence was correlated with higher basal areas of coniferous trees and components of mature forests (i.e., increased availability of snags, greater amounts of advanced decay class downed woody debris, and higher basal area of canopy trees) secondary hardwood Great

Lakes-St. Lawrence forests.

Snag availability has previously been linked to increased densities of G. sabrinus in central Ontario (Holloway and Malcolm 2006, Holloway and Malcolm 2007). Tree cavities are the preferred nesting sites for G. sabrinus (Cowan 1936, Weigl and Osgood 1974, Maser

1981) and are most commonly found in large diameter snags (Mannan et al. 1980, Rosenberg et al. 1988) in mature forests (Rosenberg and Anthony 1992). While cavity nests provide sites for raising young, avoiding predators, and improving thermoregulation (Weigl 1978), the availability of suitable cavity trees (i.e., snags) may not be limiting to the occurrence of

G. sabrinus in secondary hardwood forests. The species has proven to be quite versatile in terms of nest site selection when cavities are not readily available. For instance, the use of external stick, moss or leaf nests, as well as subterranean nests, has been well documented

(Cowan 1936, Weigl and Osgood 1974). In regions not characterised by mature forest or old- growth legacies, such as secondary hardwood Great Lakes-St. Lawrence forests, it is likely that G. sabrinus uses external or subterranean nests when appropriate snags are not readily available. Additionally, primary cavity excavators, such as woodpeckers, which secondary cavity nesters rely upon to create most tree cavities, may be less abundant or not present at all in smaller, isolated fragments of second-growth forest (Conner and Rudolph 1991). While snag availability did not appear to influence G. sabrinus occurrence, it is possible that this factor has a greater impact on G. sabrinus densities (Holloway and Malcolm 2006), which were not investigated in this study.

Conifer structure, in particular the density of spruce trees (Picea sp.), has been linked to increased densities of G. sabrinus in the Great Lakes-St. Lawrence forest region of Ontario (Holloway and Malcolm 2006). Spruce appears to be important to G. sabrinus as both direct and indirect food sources (Holloway and Malcolm 2006): the seeds may be consumed directly (although this behaviour is more typical of T. hudsonicus) and the mychorrizal fungi

(Elaphomyces sp.) most common in the diet of G. sabrinus is highly associated with spruce trees (Holloway and Malcolm 2006). Additionally, eastern white cedar (T. occidentalis) appears to be an important resource to G. sabrinus for use as nesting material in secondary hardwood Great Lakes-St. Lawrence forests (see Chapter 2). In my study, spruce trees were absent from nearly all sites (16 of 24) and, when present, only accounted for about 4% of the tree species composition, on average, by basal area. While conifer structure appears to be an important resource for G. sabrinus in many regions, I could find little evidence that the basal area of conifers had any influence on predicting the occurrence of G. sabrinus.

A main trend in the literature has been that the quality of habitat provided by old- growth forest stands is greater than that of second-growth stands for G. sabrinus. This trend has, however, been recently disputed by several studies showing that G. sabrinus abundance does not vary appreciably between old-growth and second-growth forests. Wheatley et al.

(2005) tested for relationships among broad habitat categories in Alberta, including forest age, and found that G. sabrinus was equally abundant in younger, second-growth forests and old-growth forests. Waters and Zabel (1995) and Pyare and Longland (2002) could also not find any relationship between G. sabrinus abundance and features typically considered to be characteristic of mature forests, such as the availability of cavities and snags. Additionally,

Wheatley et al. (2005) could find little correlation between G. sabrinus abundance and conifer structure. Thus, contrary to many reports that G. sabrinus is a conifer and old-growth specialist, a growing body of research is beginning to elucidate G. sabrinus as a plastic 49 habitat opportunist, capable of adjusting its biology to a wide range of habitat conditions

(Weigl 2007). The lack of statistically significant correlation between habitat variables and G. sabrinus occurrence in my study supports the findings that G. sabrinus is able to successfully occupy a wide range of habitat types, and, thus, G. sabrinus appears as a habitat generalist in secondary hardwood Great Lakes-St. Lawrence forests. My findings that G. sabrinus are not necessarily specific to coniferous-dominated habitat or features of mature forests supports the findings of Doyle (1990), Rosenberg and Anthony (1992), Martin (1994), Cotton and Parker

(2000), Ransome and Sullivan (2003), and Wheatley et al. (2005).

It is not known whether unexplored variables played a key role in governing habitat use of G. sabrinus. For instance, the occurrence of G. sabrinus may be related to food abundance (Wheatley et al. 2005, Holloway and Malcolm 2006). Glaucomys sabrinus is a mycophagist with relatively specific diet requirements of subterranean fruiting bodies of hypogeous fungi (Maser et al. 1986, Currah et al. 2000, Holloway 2006). However, a feasible methodology does not exist to relate hypogeous fungi abundance with measurable habitat features (Wheatley et al. 2005), hence identifying food abundance as a significant factor controlling G. sabrinus occurrence is difficult. Other food types are undoubtedly important as well (e.g., hard-bodied insects and fruit; Holloway 2006) and may even substitute for fungi.

It is possible that the aforementioned habitat features are density-limiting features in secondary hardwood Great Lakes-St. Lawrence forests, however this hypothesis was not tested here. There is a need for further empirical studies aimed at enhancing our understanding of G. sabrinus habitat requirements in regions where old-growth forest or old- growth legacies are not commonplace. As well, future studies of the biological mechanisms controlling fragment occupancy should seek to incorporate measures of nest requirements, food availability, and predation rates across a range of fragment sizes.

4.3 Responses ofG. sabrinus to landscape structure features

Patch area was found to be the strongest correlate of G. sabrinus occurrence in secondary hardwood Great Lakes-St. Lawrence forests. The patch area requirement threshold for G. sabrinus as judged from the receiver operating characteristic curve, was found to be

46.5 ha. This value represents the patch size at which the probability of G. sabrinus occupying a patch in secondary hardwood Great Lakes-St. Lawrence forests is maximised, while the probability of G. sabrinus not occupying a patch is minimised. I used this threshold value of patch size to map all patches in Bruce and Grey counties, Ontario at which the probability of positively identifying G. sabrinus occupancy is maximised (Figure 1.9). An approximate minimum patch size was 29 ha, in that G. sabrinus did not occur in any patches smaller than this (n = 7). Only three (59 ha, 73 ha, 92 ha) of the seventeen patches greater than 29 ha in size were unoccupied. Smith and Person (2007) identified that a 41 ha patch of upland old-growth forest habitat had an 85% probability of sustaining a simulated G. sabrinus population for 25 years in old-growth forest stands in southern Alaska. Larger patch sizes were necessary to support G. sabrinus populations for longer periods of time. For example, a minimum patch size of 1136 ha was required to support a simulated G. sabrinus population for 100 years (Smith and Person 2007). Several other species of flying squirrels have also demonstrated strong patch area-occurrence relationships. For instance, Nupp and

Swihart (2000) observed a positive relationship between southern flying squirrel (G. volans) occurrence and patch area in Indiana. The approximate minimum patch size for G. volans 51

I 450000 m E

Figure 1.9: Patch suitability map for Glaucomys sabrinus in Bruce and Grey counties, Ontario based on the threshold patch area requirement derived from the receiver operat ing characteristic (ROC) curve and logistic regression model. Black forest patches repre sent patches of 46.5 ha or greater and, thus, are patches where Glaucomys sabrinus oc currence is probable. Grey forest patches represent forest areas of 46.4 ha or smaller and, thus, are patches where Glaucomys sabrinus is not likely to occur. Two Universal Trans verse Mercator, Zone 17 grid lines also are shown (map datum: NAD83). 52 occurrence was determined by Nupp and Swihart (2000) as 4.6 ha, quite different from the 29 ha required by G. sabrinus in this study. In Europe, patch size was found to correlate with the probability of the Siberian flying squirrel (P. volans) vacating an occupied patch (Selonen and Hanski 2003). This relationship is driven by the high probability of P. volans abandoning small patches in favour of larger patches in a fragmented landscape (Selonen and Hanski

2003), presumably due to resource deficiencies in smaller patches. Furthermore, several

Eurasian tree squirrels have demonstrated a positive relationship between patch size and occurrence (Verboom and van Apeldoorn 1990, Fitzgibbon 1993, van Apeldoorn et al. 1994,

Wauters et al. 1994).

The way in which an organism responds to habitat edges is thought to predict the organism's response to matrix and corridors (Wiens et al. 1985, Stamps et al. 1987, Haddad

1999, Selonen and Hanski 2003), and thus influence an organism's ability to disperse throughout the landscape. For instance, a species that actively avoids habitat edges should not be able to effectively use either the matrix or corridors for movement between patches

(Laurance 1991, Laurance 1995). Glaucomys sabrinus individuals have been observed foraging in forest habitat edges in southern and central Ontario (pers. obs., Hollow ay 2006).

Additionally, Bayne and Hobson (1998) found no significant difference in G. sabrinus abundance between edge and interior forest habitats in an agricultural matrix. Research on P. volans in Europe using radio-telemetry has shown that this species uses corridors when present, and matrix when corridors are not present, to move between preferred patches of habitat (Selonen and Hanski 2003). Thus, from a theoretical standpoint, G. sabrinus should be able to effectively (i) use corridors for interpatch movement and (ii) travel in or through matrix when the distance between patches is not great (i.e., < 100-200 m) and some cover is 53 available (Selonen and Hanski 2003). A landscape genetics study of G. sabrinus in the Great

Lakes-St. Lawrence forest region of Ontario has shown little genetic differentiation within and amongst subpopulations in Bruce and Grey counties, Ontario and between populations throughout this region (i.e., little evidence of population structure; McEachen 2007). This finding suggests a high gene flow between individuals in a fragmented population and supports my hypothesis that G. sabrinus is able to readily move about in fragmented landscapes, possibly using matrix and corridors to achieve natal dispersal. It is possible that corridors are more important for the colonisation of isolated patches than the actual distance between patches, which may explain my finding that G. sabrinus occurrence did not correlate with distance-based measures of patch isolation. Instead, the density of possible movement corridors or least-cost pathways may better serve to highlight relationships between landscape composition and G. sabrinus occurrence (Apeldoorn et al. 1994).

By only showing sensitivity to patch area, the distribution of G. sabrinus in secondary hardwood Great Lakes-St. Lawrence forests appears to be driven by local extinctions in, or abandonment of, small patches followed by little to no recolonisation. If true, then, minimum patch size may increase with time as subpopulations "wink out" of existence. An alternative possibility is that even the most isolated patches are periodically colonised and that other factors determine minimum patch size. For example, the minimum suitable area of a forest patch for G. sabrinus may be limited by thermoregulatory requirements associated with group huddling, especially during the winter (Stapp et al. 1991, Nupp and Swihart 2000).

This life-history strategy may result in G. sabrinus being more susceptible to local extinctions in small habitat patches that cannot support the density of individuals required to sufficiently reduce their winter energy expenditures (see Stapp et al. 1991 for more 54 discussion of thermoregulatory requirements of flying squirrels). Thus, from a metapopulation perspective, local extinctions of G. sabrinus subpopulations in small habitat patches within an agricultural matrix are likely to occur more frequently than in larger patches, perhaps due to thermoregulatory pressures and other ecological factors, such as predation. Smaller patches are also less likely to be followed by recolonisation events, especially when connectivity is low, as predicted by the target-area hypothesis, which states that dispersing individuals have a greater probability of encountering larger habitat patches

(Lomolino 1990, Goodwin et al. 1999, Goheen et al. 20036).

4.4 Responses ofT. hudsonicus to habitat features

In this study T. hudsonicus occurred in 50% of the studied forest fragments; all patches in which T. hudsonicus occurred were >9 ha in area. None of the landscape structure variables were significant predictors of T. hudsonicus occurrence, indicating that, in this study, T. hudsonicus was not influenced by varying degrees of habitat fragmentation. Instead, the deciduous-coniferous tree composition, as determined by tree basal area, was the proximate factor predicting the occurrence of T. hudsonicus in secondary hardwood Great

Lakes-St. Lawrence forests. While few studies have quantified habitat selection by T. hudsonicus in predominantly deciduous forests, the individuals observed in this study seemed to prefer stands with at least some coniferous structure, consistent with other findings (Nupp and Swihart 2000, Swihart et al. 2007). For instance, a study conducted by Swihart et al.

(2007) in a deciduous-dominated, fragmented landscape in Indiana found that T. hudsonicus most frequently used habitat containing conifers and black walnut trees. T. hudsonicus are 55 conifer specialists throughout their range in North America, feeding primarily on the seeds of conifer trees such as Picea sp. and Pinus sp. and relying on conifer seeds almost exclusively for overwinter survival (Smith 1968, Kemp and Keith 1970, Rusch and Reeder 1978, Fisher et al. 2005). Despite this, T. hudsonicus is prevalent throughout decidous-dominated forests in southern Canada and the United States, suggesting that T. hudsonicus are capable of functioning reasonably as habitat generalists (Swihart and Nupp 1998). Phenotypic plasticity of behavioural traits has previously been observed in T. hudsoncius, specifically with regards to the presence or absence of territoriality (Layne 1954, Smith 1968, Rusch and Reeder

1978), larder or scatter hoarding of seeds (Layne 1954), the use of either cavity nests or external dreys (Layne 1954, Yahner 1980), and diet (Pretzlaw et al. 2006). There is little evidence, however, that T. hudsonicus uses habitat in which coniferous structure is absent or actively avoids coniferous structure (except see Goheen and Swihart 2005), suggesting that they are not true habitat generalists. While I was able to detect the presence of T. hudsonicus in deciduous-dominated forest stands, there was a strong preference towards increased basal area of coniferous trees (and, by corollary, reduced basal area of deciduous trees). Therefore, in secondary hardwood Great Lakes-St. Lawrence forests, deciduous-coniferous tree composition appears to be the proximate factor driving T. hudsonicus occurrence.

Tamiasciurus hudsonicus occurrence did not significantly correlate with any of the additional habitat structure variables measured in this study, including canopy closure, basal area of canopy and subcanopy trees, basal area of snags and amount of downed woody debris.

4.5 Responses ofT. hudsonicus to landscape structure features Contrary to some studies, I was unable to detect significant effects of habitat fragmentation on T. hudsonicus occurrence. In the midwestern United States, T. hudsonicus occurrence has been linked with increasing patch area and reduced amounts of edge habitat

(Nupp and Swihart 2000, Goheen et al. 2003a). Bayne and Hobson (2000) also discovered a negative relationship, albeit a weak one, between T. hudsonicus abundance and patch size, with highest abundance in small forest patches in an agricultural matrix. The Eurasian red squirrel (Sciurus vulgaris) rarely crosses open fields (Andren and Delin 1994), a behaviour which may limit its dispersal and colonisation abilities; however, there is little evidence from radio-telemetry studies to support a similar behaviour in T. hudsoncius. In fact, T. hudsoncius has managed to successfully expand its range in the midwestern United States concurrent with agriculturally induced fragmentation of deciduous-dominated forests and widespread afforestation of conifers (Goheen et al. 2003a, Swihart et al. 2007). In stochastic simulation modelling of red squirrel metapopulation persistence in Indiana, T. hudsonicus had the longest time to extinction of the four tree squirrel species studied (compared to fox [Sciurus niger], southern flying [G. volans], and grey [Sciurus carolinensis] squirrels; Swihart and

Nupp 1998). According to Swihart and Nupp (1998), the effects of habitat fragmentation on

T. hudsonicus are buffered by the large population growth potential of the species and a well- developed dispersal ability, which permits rapid recolonisation of unoccupied patches. These factors may be invoked to explain my finding that T. hudsonicus occurrence in a fragmented deciduous-dominated landscape is not strongly influenced by habitat fragmentation.

The variation in results among habitat fragmentation studies suggests that T. hudsonicus exhibits some plasticity in their response to landscape structure. This variation also serves to highlight the inherent problems in assigning universal importance to context- 57 specific variables, such as landscape structure, based on studies conducted at one spatial scale or region. Differential responses to habitat fragmentation are likely a consequence of variations in the patch context in which a species occurs. Therefore, difficulties may arise when predicting species responses to landscape change in one region by extrapolating findings from another. Conceptually, the variation in T. hudsonicus response to habitat fragmentation could be due to divergent metapopulation dynamics that are driven by differences in a landscape's configuration, composition, and connectivity of suitable habitat

(Vos et al. 2001, Swihart et al. 2007). Metapopulation studies would benefit immensely from comparative analyses of the distribution and patch occupancy rates of single species in landscapes with various configurations and compositions of habitat (sensu Andren 1994).

4.6 Misclassifying absence

Error due to non-detection of species occurrence may have reduced the explanatory power of my models (MacKenzie 2005). Non-detection of species occurrence in an occupied patch is a common problem in presence-absence based studies, especially when the population size is small, as may be expected in highly fragmented landscapes (Andren 1994).

Non-detection has been shown through both field-based and simulation studies to yield biases in parameter estimation of logistic regression models (Gu and Swihart 2004). I tried to minimise such bias by establishing a pre-bait period prior to each trapping session and by trapping the same sites on two separate occasions in the same season. I found that the mean number of trap nights to detection of G. sabrinus and T. hudsonicus was 79.6 and 60.1 trap nights, respectively. Thus, detection probabilities for G. sabrinus and T. hudsonicus were 58

0.63 and 0.72, respectively, suggesting that each species was detected in a patch in much fewer than the total (n = 216) number of trap nights on average (Figure 1.10). A detection probability equal to 1 represents the case where a species was detected without any effort and detection probability of 0 represents the case where a species is detected, in this case, on the final night of trapping.

4.7 Future directions and conclusions

The objective of this study was to test the observed patterns of G. sabrinus and T. hudsonicus occurrence to habitat and landscape features and interpret the results to an extent permitted by the ecological literature. The investigation and relative importance of biological mechanisms, while not tested here, will be particularly valuable in future research. Topics of specific interest to researchers should be: the dispersal abilities of sciurids in fragmented landscapes, particularly G. sabrinus, and their utilisation of matrix and corridors; developing a clear understanding of the metapopulation dynamics (i.e., extinction and colonisation rates) for both G. sabrinus and T. hudsonicus; the role of niche partitioning and interspecific competition in structuring sciurid communities in fragmented landscapes that include both G. sabrinus and T. hudsonicus; and, the relationship between G. sabrinus presence-absence and mycorrhizal fungi availability as a possible indicator of habitat quality. Additional investigation of G. sabrinus habitat requirements in regions not characterised by old-growth coniferous forest or mature forest legacy features is also required. Continued investigation of

T. hudsonicus populations in deciduous-dominated landscapes may lead to a greater understanding of their habitat requirements in southern Canada and the eastern United States. 12

• Glaucomys sabrinus • Tamiasciurus hudsonicus

No. of nights to first detection

Figure 1.10: Frequency distribution of the number of fragments in which occurrence was deter mined as a function of time to first detection (detection probability) for Glaucomys sabrinus and Tamiasciurus hudsonicus in a fragmented secondary hardwood Great Lakes-St. Lawrence forest region. Studies of metapopulations must consider the temporal nature of population responses and, as such, should be carried out over long time frames to better reveal the mechanisms governing the patterns and processes under investigation. While commuting landscape structure-species occurrence results between regions or landscape contexts currently remains a difficult task, such goals should be at the forefront of future landscape ecology studies. As such, I must caution against extrapolating the results of this study to other regions, especially those characterised by contrasting land uses and forest compositions. CHAPTER 2: CAVITY NEST MATERIAL USE BY NORTHERN FLYING SQUIRRELS AND RED

SQUIRRELS IN SOUTHERN ONTARIO: A CASE FOR THE NEST-PROTECTION HYPOTHESIS

ABSTRACT

Despite the importance of tree cavities to arboreal rodents, their use as nesting habitat is poorly understood. Through deployment of artificial nest boxes, I examined nest materials used by northern flying squirrels (Glaucomys sabrinus) and red squirrels (Tamiasciurus hudsonicus) in a secondary hardwood Great Lakes-St. Lawrence forest region. I collected

224 nests between 2002 and 2005 and found that 83% were constructed almost entirely of shredded bark from eastern white cedar {Thuja occidentalh). Mean nest depth was 9.5 cm and showed no difference between species or between spring and summer nests. However, nest depth did vary with the type of material used, with nests containing cedar bark were significantly deeper than those without (means were 9.9 cm and 7.5 cm, respectively), suggesting the existence of a trade-off between the thermoregulatory benefits and antiparasitic effects of eastern white cedar bark. I review the antiparasitic properties of eastern white cedar and suggest that the use of shredded cedar bark by northern flying squirrels and red squirrels to line nest cavities is a behavioural adaptation to reduce ectoparasite loads.

61 62

1. INTRODUCTION

Tree cavities are a critical resource for many vertebrates in forest ecosystems, providing sites for raising young, avoiding predators, and improved thermoregulation

(Collias 1964, Wiebe 2001). Tree cavities are generally most abundant in old growth forests where large diameter trees and well-decayed snags are commonplace (Holloway and

Malcolm 2007, Holloway and Malcolm 2006). Cavities also may form in live trees where the entry of rot has been facilitated by disease, deformities, broken limbs or woodpecker excavations. Where tree cavities are not readily available, many members of the family

Sciuridae use external nests (dreys) or subterranean nests, although evidence is emerging that many arboreal vertebrates preferentially select tree cavities when they are available (Bakker and Hastings 2002).

The nocturnal northern flying squirrel (Glaucomys sabrinus) and the diurnal red squirrel {Tamiasciurus hudsonicus) are cavity nesting, arboreal rodents common in temperate and boreal forests of North America. Glaucomys sabrinus is a small, gliding mammal that has been rarely studied in the eastern portion of its range (but see Holloway and Malcolm

2007, Holloway and Malcolm 2006). Glaucomys sabrinus is considered to be a keystone species in Pacific Northwest forest ecosystems, where it serves as an important prey species of the northern spotted owl (Strix occidentalis) and is thought to play an essential role in the dissemination of mycorrhizal fungi spores (Carey 2000). Tamiasciurus hudsonicus is somewhat larger in size than G. sabrinus and is more adept at ground-based locomotion than

G. sabrinus. Both species are primarily associated with conifer-dominated forests, although in some areas of eastern North America they inhabit hardwood and mixed hardwood-conifer 63 forests (Flyger and Gates 1982, Holloway and Malcolm 2006, Nupp and Swihart 2000).

Glaucomys sabrinus and T. hudsonicus rely on similar behavioural strategies for conserving energy in the winter months, including food caching, reduced foraging activity, and nesting in tree cavities (Layne 1954, Rothwell 1979, Cotton and Parker 2000). Flying squirrels are unique in that they can form large winter aggregations, which are thought to increase internal cavity temperatures and, hence, reduce individual energy expenditures (Contreras 1984,

Stapp et al. 1991).

Nest substrates used by arboreal squirrels, including the southern flying squirrel

{Glaucomys volans), have been found to vary geographically (Muul 1974). However, most reports describing nest materials employed by squirrels have been qualitative in nature and have focused on dreys. Tamiasciurus hudsonicus dreys have been noted to include grape bark, deciduous leaves, dried grasses, moss, feathers, fur and soft inner bark (Layne 1954).

Similarly, dreys of G. sabrinus have been found to consist of dried grasses, shredded bark, mosses, twigs and lichens (Rust 1946, Mowrey and Zasada 1984). Cavity nests have been described quantitatively only once for G. sabrinus (Hayward and Rosentreter 1994). Using nest boxes, these authors found that G. sabrinus in the northern Rocky Mountains of central

Idaho and western Montana primarily constructed cavity nests from arboreal lichens, especially of the genera Bryoria and Letharia. I am not aware of studies that have quantified cavity nest materials of T. hudsonicus; however, Hayward and Rosentreter (1994) suggest a preference for grasses and shredded bark. Published studies of G. sabrinus and T. hudsonicus nest materials in the Great Lakes-St. Lawrence forest region of eastern North America apparently do not exist; similarly, studies on seasonal differences in nest materials and nest depths have not been undertaken. Despite their importance, cavity nests appear to be rarely incorporated into studies of arboreal vertebrates (Fokidis and Risch 2005), in part because of difficulties of access.

Increasingly, nest boxes are being used as a means by which to study and monitor cavity nesters. While other methods (e.g., radio-telemetry) do exist, difficulties in systematically sampling and capturing individuals from within cavities has limited the widespread adoption of such techniques for studying nesting behaviour or for monitoring population status and reproductive parameters. Nest boxes have been used successfully to study the nesting behaviours and nesting materials of G. sabrinus (Hayward and Rosentreter 1994), G. volans

(Layne and Raymond 1994), as well as many other species of mammals and birds (Havelka and Millar 1996, Male et al. 2006).

The purpose of this study was to characterise cavity nest materials used by G. sabrinus and T. hudsonicus in a large network of nest boxes in the Huron Region of the Great

Lakes-St. Lawrence forest. Nest depth measurements were investigated as a function of the type of nest material and season (spring and summer); I also looked for seasonal variation in nest materials. Because of the prevalence of shredded cedar bark in the nests of both G. sabrinus and T. hudsonicus, I speculate on the potential antiparasitic benefits of eastern white cedar {Thuja occidentalis) bark.

2. METHODS

A total of 266 nest boxes were installed between 2002 and 2004 as part of a larger study being conducted on G. sabrinus. Sampling effort varied across years due to logistical constraints. In 2002, 154 boxes were installed; an additional 36 boxes were installed in 2003, 65 and 76 boxes were installed in 2004. All nest boxes were placed at heights of 3-4.5 m on trees that had a diameter at breast height >17 cm. Following Carey (2002), nest boxes were established at a density of three nest boxes/ha to allow for a balance between occupancy rates and monitoring effort. Three or more boxes were placed at each of 70 sites across the study region.

Boxes were constructed from unfinished, 3-cm thick douglas fir (Pseudotsuga menziesii) lumber following Sonenshine et al. (1973). Our nest box design differed slightly from Sonenshine et al. (1973) in that I did not include a hinged door or wire screen at the front of the shelter. Instead, our boxes were constructed such that the top lid could be flipped up to remove the animals. Our entrance hole was also of larger diameter (4 cm) to permit use by G. sabrinus and T. hudsonicus, but exclude eastern grey squirrels (Sciurus carolinensis).

All boxes were screwed to the tree trunk and any tree branches were removed from below the nest box. Because a main goal of this particular study was to document nest material, I did not place initial nest material in the boxes.

All boxes were checked twice per year during the day: once in the spring and again in autumn. Spring checks were conducted between May 12 and June 12 following the onset of parturition, whereas autumn checks were conducted between October 12 and November 12.

Preliminary monthly nest box searches indicated that both G. sabrinus and T. hudsonicus did not use the nest boxes in late autumn and winter (late November - late February). Therefore, nests found at the spring inspection had been constructed between early March and the date of inspection in spring (hereafter termed "spring" nests), whereas those found at the autumn inspection had been constructed between the spring and autumn inspection dates (hereafter termed "summer" nests). Where nests were present, I recorded the location, number and 66 reproductive condition of occupants, nest material composition and nest depth. Samples of nest material were collected from all nests for identification of materials. In cases where the nest box was not occupied at the time of inspection, the entire nest was removed. In cases where the nest box was occupied by young at the time of initial inspection (14% of the nests), the nest was left intact and subsequently removed during targeted monthly nest box examinations following the departure of nest occupants. I removed nest material not only to examine nest materials, but also because Hayward and Rosentreter (1994) observed that nest boxes were typically not occupied in successive years if used nest material was present. Nest depth was always determined during the initial investigation and measured in situ along the vertical dimension from the internal base of the nest box to the top of the nest. Nest volume was determined by multiplying the nest depth by the internal width and length dimensions of the nest box.

On two occasions I found deer mice (Peromyscus maniculatus) using the nest boxes.

The possibility exists that some P. maniculatus nests from unoccupied nest boxes may have been included in our analysed sample; however, the low occurrence of P. maniculatus (6%) lead us to infer that these occurrences were rare and that the vast majority of the unoccupied nests I found were sciurid nests. Known P. maniculatus nests were excluded from our analyses.

3. Results

3.1 Nest Box Occupation 67

Between 2002 and 2005,1 found 224 nests. Of these, 32 were occupied when checked

(14 G. sabrinus and 18 T. hudsonicus). Average litter size was 3.9 for G. sabrinus (n = 10 nests) and 4.9 for T. hudsonicus (n = 11 nests). Lone males also were found occupying the nest boxes (2 G. sabrinus and 6 T. hudsonicus).

3.2 Nest Materials ofG. sabrinus and T. hudsonicus

Shredded cedar bark was found in 186 (83%) of the nests; in nearly all of these instances, it accounted for >75% of the nest material by volume. In total, 134 (59%) of the nests were made exclusively from shredded cedar bark (Figure 2.1). Deciduous leaves (n =

8), grasses (n = 8), and moss (Sphagnum sp.) (n = 15) were the most common associates of shredded cedar bark nests.

Other materials (twigs, feathers, cedar leaves, and animal hair) also were found in conjunction with 21 (9%) of the shredded cedar bark nests. Of the 224 nests, 24 (11%) were composed of >50% deciduous leaves by volume. In 16 (7%) of these cases, deciduous leaves made up 100% of the nest material by volume. Trace amounts of dried grasses, shredded plastic, white pine needles, and peat moss where found in 9 (4%) of the deciduous leaf dominated nests. Grass-dominated nests represented 6 (3%) of the samples, with other materials, such as peat moss (n = 3), deciduous bark (n = 3), and fibreglass insulation (n = 2) dominating fewer nests.

In boxes occupied by either G. sabrinus or T. hudsonicus, the materials used did not appear to differ from the above characterisation of all nest samples. Twelve of the 14 boxes

(86%) occupied by G. sabrinus and 13 of the 18 boxes (72%) occupied by T. hudsonicus 68

Cedar bark Deciduous Grasses Sphagnum Deciduous Fibreglass leaves moss bark insulation Nest material Figure 2.1: Total number of cavity nests from the complete sample (n = 224) (i.e., known and unknown occupants) that were dominated (> 50% by volume) by the various materials in a secondary hardwood Great Lakes-St. Lawrence forest region 69 contained >90% shredded cedar bark by volume (Figure 2.2). One G. sabrinus nest was composed entirely of fibreglass insulation and another was composed of >50% shredded deciduous bark. Trace amounts of peat moss, dried grasses, cedar leaves, and twigs were found in occupied G. sabrinus nests. Two T. hudsonicus nests were composed entirely of dried grasses and one nest contained >80% deciduous leaves. Trace amounts of shredded plastic, bird feathers, animal fur, dried grasses, and deciduous leaves also were found in occupied T. hudsonicus nests. Nest materials did not appear to differ between seasons, with shredded cedar bark dominating nests in both spring and summer (t = -1.892, p = 0.06).

Shredded cedar bark was found to occur in 84.5% of spring nests (n = 164) and 79.3% of summer nests (n = 24).

3.3 Nest Depth

Mean nest depth across the 224 samples was 9.5 cm (SD = 4.4, range = 1.3 to 18.3 cm). Mean nest depth of nests occupied by G. sabrinus was 12.1 cm (SD = 2.4, range = 7.7 to 15.3 cm) and for T. hudsonicus was 12.3 cm (SD = 2.1, range = 10.1 to 17.8 cm), although the difference was not statistically different {t = 0.31, p = 0.76). However, nest depth varied according to nest material: nests that contained cedar bark were significantly deeper than those that did not {t = 3.20, p = 0.002; respective means were 9.9 cm [n = 168, SD = 1.7] and

7.5 cm [n = 36, SD = 1.5]). I found no statistical difference between spring and summer nest depths for the combined sample (t = -1.892, p = 0.128) or for either species for occupied nests (G. sabrinus: t = 0.55,p = 0.30; T. hudsonicus: t = 0.ll,p = 0.43). Mean spring nest depth was 9.5 cm (n = 194) and mean fall nest depth was 8.9 cm (n = 30). 70

\Glaucomys sabrinus I Tamiasciurus hudsonicus

Cedar bark Deciduous Grasses Deciduous Fibreglass leaves bark insulation Nest material Figure 2.2: Species-specific distribution of dominant (> 50%) nest materials found in nest boxes known to be occupied by Glaucomys sabrinus (n = 14) and Tamiasciurus hudsonicus (n = 18) in a secondary hardwood Great Lakes-St. Lawrence forest region 71

4. DISCUSSION

Nest materials found in nest boxes used by G. sabrinus and T. hudsonicus were usually composed almost entirely of shredded bark from eastern white cedar. The use of shredded cedar bark was disproportionate to the abundance of other potential nest materials, such as deciduous leaves (11% of nests) and dried grasses (3% of nests).

The prominence of cedar bark in G. sabrinus and T. hudsonicus nests may be explained by at least two, non-mutually exclusive hypotheses: nest-protection and thermoregulation. The "nest-protection hypothesis" has received much attention from ornithologists to explain the use of green vegetation in avian nests. Initially suggested by

Wimberger (1984) and tested by Clark and Mason (1985), the nest-protection hypothesis posits that birds exploit the antiparasitic properties of certain plant species that emit volatile compounds (Dawson 2004). All plants contain secondary metabolites that are used as a defense against disease and herbivory (Clark and Mason 1988) and when used by birds these compounds may reduce ectoparasite loads in the nest environment. Numerous studies have supported the nest-protection hypothesis in birds, although most research has been on the

European starling (Sturnus vulgaris L.) (Wimberger 1984, Rodgers et al. 1988, Fauth et al.

1991, LaFuma et al. 2001). In field and laboratory tests, the volatile compounds of wild carrot (Daucus carota L.) and fleabane (Erigeron philadelphicus L.) reduced the number and emergence of hematophagous mites in S. vulgaris nests (Clark and Mason 1988). Further, in a review of the composition of nest material for 137 passerine birds breeding in eastern North

America, Clark and Mason (1985) found that passerine species nesting in enclosed spaces

(i.e., secondary cavities or nest boxes) were more likely to construct their nests from volatile compound-producing plants in comparison to species that had open-cup nests. This finding is particularly relevant because species that burrow or nest in enclosed spaces may be particularly susceptible to the accrual of nest-borne ectoparasites due to favourable microclimatic conditions (Hemmes et al. 2002). The nestprotection hypothesis has found support on at least one occasion for small mammals: Hemmes et al. (2002) reported that dusky-footed wood rats (Neotoma fuscipes Baird) placed California bay (Umbellularia californica Nuttall) leaves around their nest sites. Laboratory tests revealed that when incubated with torn U. californica leaves for 72 h, survival of flea larvae was reduced by

74% compared to controls.

By use of shredded eastern white cedar bark as a primary nest material, G. sabrinus and T. hudsonicus also may be limiting exposure to nest-borne ectoparasites. Phytochemical analysis of T. occidentalis bark revealed 22 volatile compounds, including: monoterpenes, fenchene, camphene, camphor, carvacol, and paracymene (Shaw 1953, Witte et al. 1983,

Yatagai et al. 1985, Keita et al. 2001). Keita et al. (2001) investigated the insecticidal effects of these 22 volatile compounds (primarily a-thujone) and noted several negative biological effects on the cowpea weevil (Callosobruchus maculatus Fabricius), including complete suppression of egg laying by adult females, near complete reduction in egg hatching (98.8%), and 39% to 98% mortality in adults. Additionally, several studies have shown that the essential oils derived from Juniperus excelsa, another member of the Cupressaceae family, have potent insecticidal properties (Adams et al. 1988, Adams 1993).

The nest protection hypothesis also gains support from the use of lichens in western nests of G. sabrinus (Hayward and Rosentreter 1994). Antiparasitic effects have been described for several species of lichen; in particular, usnic acid (a yellow cortical pigment 73 found only in lichens) exhibits antipathogenic, anti-inflammatory, and analgesic activities in mammals (Al-Bekairi et al. 1991, Lauterwein et al. 1995, Okuyama et al. 1995, Ingolfsdottir et al. 1998). Recently, De Carvalho et al. (2005) reported on the inhibitive activities of usnic acid on the growth and viability of the parasitic protozoan Trypanosoma cruzi, which is transmitted to humans and other mammals by triatomine bugs (Triatoma sp.). Lichens belonging to the genus Usnea contain high quantities of usnic acid (Nishtoba et al. 1987); however, these species are not often used as nest material by G. sabrinus in the western boreal forest, perhaps due to limited availability (Hayward and Rosentreter 1994). Instead, G. sabrinus selected lichens containing the secondary compounds norstictic acid, fumarprotocetraric acid, vulpinic acid, and atranorin, all of which have been attributed to varying degrees of antiparasitic or antimicrobial effects (Tay et al. 2004, Yilmaz et al. 2004,

Giez et al. 1994). These various studies support the possibility that the use of T. occidentalis, as well as several lichen species, by G. sabrinus and T. hudsonicus may be a behavioural adaptation for parasite control in the nest environment. Further testing of the antiparasitic effects of T. occidentalis volatile oils and other secondary compounds found in G. sabrinus and T. hudsonicus nesting materials is required.

G. sabrinus and T. hudsonicus are host to several ectoparasites, including some that are considered to be endemic to each species. Some of the ectoparasites commonly found in association with G. sabrinus include: several flea species (such as Opisodasys pseudarctomys

Baker); a sucking louse, Hoplopleura trispinosa (Kellogg and Ferris) (Pung et al. 2000); and a species of tick, Ixodes angustus (Neumann) (Murrell et al. 2003). Tamiasciurus hudsonicus has been shown to host six species of parasitic mites, including Dermanyssus prognephilus

(Ewing); a chigger species (Euschongastia setosa Ewing); one species of tick {Ixodes marxi Banks); and eleven species of fleas, including Epitedia wenmanni (Rothschild), Monopsyllus vison (Baker), and Orchopeas howardii (Baker) (Layne 1954).

A second hypothesis is that shredded cedar bark may offer greater insulative properties than other available materials. Hay ward and Rosentreter (1994) suggested the same hypothesis when characterising the primary use of lichens as nest material by G. sabrinus. This hypothesis is based on Stapp et al.'s (1991) finding that nests composed of plant fibres allow southern flying squirrels to reduce their energy expenditure when experimentally subjected to cold temperatures. I am not aware of studies on the thermal properties of cedar bark, lichens or other nest materials found in this study. I found that shredded cedar bark was used equally in spring and summer/early fall and that nest depth did not vary between the two seasons. A common assumption is that nest materials with greater thermal properties are favoured, and/or that nest depth would peak, during colder periods.

Inspection during additional months would better test this possibility given that ambient temperatures did not contrast strongly between our two sampling periods. Sample sizes for known female and male nests were too small to allow for formal testing of hypotheses; however, these samples provide some indication that males of both G. sabrinus and T. hudsonicus construct nests that are much shallower than females. Young, furless squirrel pups are at an increased risk of energy-related mortality, which may help to explain why females appear to invest more time and energy into constructing larger nests than males.

While modest work has been accomplished on the potential thermoregulatory benefits of nest materials used primarily by birds (Hilton et al. 2004), further analysis of specific nest materials used by G. sabrinus, T. hudsonicus, and other small mammal species is required. 75

Nest depth was statistically correlated with the type of nest material used, such that nests containing at least some shredded cedar bark were significantly deeper than nests that did not contain any shredded cedar bark. One hypothesis is that the use of cedar bark involves a tradeoff between antiparasitic and thermal properties. Perhaps squirrels favour cedar due to its antiparasitic properties, but construct thicker nests in this case because of its poorer insulative properties compared to other potential nesting materials. Alternatively, the structural characteristics of deciduous leaves and dried grasses may make them more susceptible than cedar bark to compaction. 76

GENERAL CONCLUSIONS

In this study, Glaucomys sabrinus demonstrated quite flexible habitat requirements and little evidence of a response to degrees of either patch isolation or landscape composition, although patch size was an important fragment feature. Provided that tracts of forest greater than 46.5 ha are available and retained in the landscape, and given some 20% or greater forest cover, suitable habitat for G. sabrinus appears to be widely available in secondary hardwood Great Lakes-St. Lawrence forests. I speculate that G. sabrinus experiences greater rates of extinction in small patches compared to larger patches owing to:

(0 thermoregulatory pressures in winter months and (ii) little to no recolonisation of patches following local extinction events, which is presumably attributable to low landscape connectivity (i.e., few suitable corridors) in this study region. I obtained evidence that eastern white cedar (Thuja occidentalis) is an important nesting resource for G. sabrinus; however, contrary to previous studies, I could find little indication that the basal area of conifers was important in explaining the occurrence of G. sabrinus.

By contrast, Tamiasciurus hudsonicus in this study were associated with relatively high conifer basal area (and, conversely, reduced basal area of deciduous trees). Therefore, conifer seed (a primary food resource) and eastern white cedar bark (a primary nesting material) may be the strongest ecological factors influencing T. hudsonicus occurrence in fragmented secondary hardwood Great Lakes-St. Lawrence forests. Similar to past findings,

T. hudsonicus did not appear to respond to landscape variables, presumably due to high vagility and high population growth potential (Swihart and Nupp 1998, Nupp and Swihart

2000, Goheen et al. 2003a). Researchers have previously noted that T. hudsonicus has 77 successfully expanded its range in the midwestern United States concurrent with increased agricultural fragmentation of deciduous-dominated forests and afforestation of coniferous trees (Swihart and Nupp 1998).

Shredded T. occidentalis bark was the predominant nest material found in nest boxes occupied by G. sabrinus and T. hudsonicus. I hypothesise, based on Clark's nest-protection hypothesis, that this represents a behavioural adaptation to reduce ectoparasites in the nest environment. Nest depth was thicker for cedar bark than for other materials, raising the possibility of compensation for its poorer insulative properties. There is a need for further investigation into the thermal properties of nest materials, the biological effects of various nest materials on ectoparasites, and the role of elevation, geography, and forest type in determining nest depths and materials.

I suggest that future research directions be aimed at determining the dispersal abilities and metapopulation dynamics of sciurids in fragmented landscapes to improve management strategies for these ecologically important species, especially for G. sabrinus, which showed a strong patch area response. One hypothesis concerning a strong patch area effect, but little evidence of isolation effects, is that G. sabrinus readily uses matrix habitat and available linear landscape features, such as corridors, to disperse between patches regardless of actual inter-patch distances. Euclidean-based measures of patch isolation do not adequately characterise landscape connectivity, instead connectivity depends on the combined effects of inter-patch distance, the nature of the matrix, the presence of suitable corridors, and species- specific responses to these landscape features. Thus, from a management perspective, corridors and stepping stone patches might prove to be important in fragmented landscapes in that they should enhance interpatch dispersal and patch recolonisation, promote genetic 78 diversity, and encourage metapopulation persistence. Consistent with other studies, I suggest that, based on the relationship between occupancy and habitat area, the preservation of large, contiguous tracts of forest be a primary goal of G. sabrinus conservation in highly fragmented regions (Smith and Person 2007). Tamiasciurus hudsonicus may derive benefits from continued afforestation of conifers in deciduous-dominated landscapes. Until more research can be conducted on sciurid habitat requirements in deciduous-dominated fragmented landscapes, old-growth legacies, such as snags, cavity trees, and advanced decay class downed woody debris, should be conserved in managed forests as a precautionary measure, where possible. At the very least, these habitats are important for myriad other species. If the absence of suitable nest sites limits some populations of obligate cavity nesters, especially secondary cavity-nesting species (Newton 1994, Bock and Fleck 1995,

Holt and Martin 1997), nest boxes may prove to be a valuable means for increasing habitat suitability.

Although G. sabrinus and T. hudsonicus appear to be quite opportunistic with respect to their habitat requirements, habitat quality can nonetheless be influenced by differing approaches to forest and resource management. Harvests of older growth forests, removals of old-growth legacy features, replacements of mature forests with tree plantations or agricultural land uses, and the increasing fragmentation of much of the remaining forest habitat all pose challenges to sciurid conservation. The survival of both species examined here is, thus, critically dependent on an improved understanding of each species' ecology, and, perhaps even more important, an awareness of the impact of human activity on their respective ecological requirements. Despite an increasing recognition of the negative effects of habitat fragmentation on a multitude of species worldwide, it is apparent that we still lack 79 pertinent knowledge of species-specific responses to varying landscape contexts and the underlying biological mechanisms. The contrasting responses of the two species investigated here suggests that conservation of individual species in fragmented landscapes will require eclectic management techniques, a wide breadth of policy instruments, and a continuation of concerted research efforts. LITERATURE CITED

Adams, R. P., M. C. McDaniel, and R R. Carter. 1988. Termiticidal activities in the

heartwood, bark/sapwood and leaves of Juniperus species from the United States.

Biochemical Systematics and Ecology 16: 453-456.

Adams, R. P. 1993. Cedar Wood Oil -Analysis and Properties. Pp. 159-173 in Essential Oils

and Waxes (H. E Linskens and J. F. Jackson, eds.). Springer-Verlag, New York, New

York.

Addicott, J. E, J. M. Aho, M. E Antolin, D. K. Padilla, J. S. Richardson, and D. A. Soluk.

1987. Ecological neighborhoods: scaling environmental patterns. Oikos 49:

340-346.

Al-Bekairi, A. M., S. Qureshi, M. A. Chaudhry, D. R. Krishna, and A. H. Shah. 1991.

Mitodepressive, clastogenic and biochemical effects of (C)-usnic acid in mice.

Journal of Ethnopharmacology 33: 217-220.

Allen, G. M., P. E J. Eagles, and S. D. Price. 1990. Conserving Carolinian Canada.

University of Waterloo Press, Waterloo, Ontario, Canada.

Andren, H., and A. Delin 1994. Habitat selection in the Eurasian red squirrel, Sciurus

vulgaris, in relation to forest fragmentation. Oikos 70: 43-48.

Andren, H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with

different proportions of suitable habitat: a review. Oikos 71: 355-366.

Andren, H. 1995. Effects of landscape composition on predation rates at habitat edges. Pp.

225-255 in Mosaic Landscapes and Ecological Processes (L. Hansson, L. Fahrig, and

G. Merriam, eds.). Chapman and Hall, London, UK.

80 81

Apeldoorn, R. C, C. Celada, and W. Nieuwenhuizen. 1994. Distribution and dynamics of the

red squirrel (Sciurus vulgaris L.) in a landscape with fragmented habitat. Landscape

Ecology 9:227-235.

Arthur, S. M., W. B. Krohn, and J. R. Gilbert. 1989. Habitat use and diet of fishers. Journal of

Wildlife Management 53: 680-688.

Bakker, V. J., and K. Hastings. 2002. Den trees used by northern flying squirrels (Glaucomys

sabrinus) in southeastern Alaska. Canadian Journal of Zoology 80: 1623-1633.

Bayne, E. M., and K. A. Hobson. 1998. The effects of habitat fragmentation by forestry and

agriculture on the abundance of small mammals in the southern boreal mixedwood

forest. Canadian Journal of Zoology 76: 62-69.

Bayne, E. M., and K. A. Hobson. 2000. Relative use of contiguous and fragmented boreal

forest by red squirrels (Tamiasciurus hudsonicus). Canadian Journal of Zoology 78:

359-365.

Bendel, R R., and J. E. Gates. 1987. Home range and microhabitat partitioning of the

southern flying squirrel {Glaucomys volans). Journal of Mammalogy 68: 243-255.

Bender, D., Contreras, T., and Fahrig, L. 1998. Habitat loss and population decline: a meta

analysis of the patch size effect. Ecology 79: 517-533.

Berteaux, D., and S. Boutin. 2000. Breeding dispersal in female North American red

squirrels. Ecology 81: 1311-1326.

Bock, C. E., and D. C. Fleck. 1995. Avian response to nest box addition in two forests of the

Colorado Front Range. Journal of Field Ornithology 66: 352-362.

Bowman, J., J. A. G. Jaeger, and L. Fahrig. 2002. Dispersal distance of mammals is

proportional to home range size. Ecology 83: 2049-2055. 82

Burnham, K. P., and D. R. Anderson, 2002. Model Selection and Multimodel Inference: A

Practical-Theoretic Approach, 2nd ed. Springer-Verlag, New York, New York.

Cale, P. G., and R. J. Hobbs. 1994. Landscape heterogeneity indices: Problems of scale and

applicability, with particular reference to animal habitat description. Pacific

Conservation Biology 1: 183-193.

Carey, A.B. 1989. Wildlife associated with old-growth forests in the Pacific Northwest.

Natural Areas Journal 9: 151-162.

Carey, A. B. 1991. The biology of arboreal rodents in Douglas-fir forests. USDA Forest

Service General Technical Report PNW-GTR-276. Pp. 46.

Carey, A. B., S. P. Horton, and B. L. Biswell. 1992. Northern spotted owls: influence of prey

base and landscape character. Ecological Monographs 62: 223-250.

Carey, A. B. 1995. Sciurids in Pacific Northwest managed and unthinned forests. Ecological

Applications 5: 648-661.

Carey, A. B., T. M. Wilson, C. C. Maguire, and B. L. Biswell. 1997. Dens of northern flying

squirrels in the Pacific Northwest. Journal of Wildlife Management 61: 684-699.

Carey, A. B., B. Kershner, B. Biswell, and L. Dominguez de Toledo. 1999. Ecological scale

and forest development: squirrels, dietary fungi, and vascular plants in managed and

unmanaged forests. Wildlife Monographs 142: 1-71.

Carey, A. B. 2000. Ecology of northern flying squirrels: implications for ecosystem

management in the Pacific Northwest, USA. Pp. 45-61 in Biology of Gliding

Mammals (R. L. Goldingay and J. S. Scheibe, eds.). Filander Verlag, Furth, Germany.

Carey, A. B. 2002. Response of northern flying squirrels to supplementary dens. Wildlife

Society Bulletin 30: 547-556. 83

Clark, L., and J. R. Mason. 1985. Use of nest material as insecticidal and anti-pathogenic

agents by the European starling. Oecologia 67: 169-176.

Clark, L., and J. R. Mason. 1988. Effect of biologically active plants used as nest material

and the derived benefit to starling nestlings. Oecologia 77: 174-180.

Collias, N. E. 1964. The evolution of nests and nest building in birds. American Zoologist 4:

175-190.

Contreras, L. C. 1984. Bioenergetics of huddling: test of a psycho-physiological hypothesis.

Journal of Mammalogy 65: 256-261.

Conner, R. N., and D. C. Rudolph. 1991. Forest habitat loss, fragmentation, and red-

cockaded woodpecker populations. Wilson Bulletin 103: 446-457.

Cotton, C. L., and K. L. Parker. 2000. Winter activity patterns of northern flying squirrels in

subboreal forests. Canadian Journal of Zoology 78: 1896-1901.

Cowan, I. M. 1936. Nesting habits of the flying squirrel Glaucomys sabrinus. Journal of

Mammalogy 17: 58-60.

Davidson, C. 1998. Issues in measuring landscape fragmentation. Wildlife Society Bulletin

26: 32-37.

Dawson, R. D. 2004. Does fresh vegetation protect avian nests from ectoparasites? An

experiment with tree swallows. Canadian Journal of Zoology 82: 1005-1010.

De Carvalho, E. A. B., P. P. Andrade, N. H. Silva, E. C. Pereira, and R. C. B. Q. Figueiredo.

2005. Effect of usnic acid from the lichen Cladonia substellata on Trypanosoma cruzi

in vitro: an ultrastructural study. Micron 36: 155-161.

De Gruchy, M. A., R. J. Reader, and D. W Larson. 2005. Biomass, productivity, and

dominance of alien plants: a multihabitat study in a national park. Ecology 86: 84

1259-1266.

D'Eon, R. G., S. M. Glenn, I. Parfitt, and M-J Fortin. 2002. Landscape connectivity as a

function of scale and organism vagility in a real forested landscape. Conservation

Ecology 6:10 [online] URL: http://www.consecol.org/vol6/iss2/artl0.

Doyle, A. T. 1990. Use of riparian and upland habitats by small mammals. Journal of

Mammalogy 71: 14-23.

Dungan, J. L., J. N. Perry, M. R. T. Dale, P. Legendre, S. Citron-Pousty, M-J. Fortin, A.

Jakomulska, M. Miriti, and M. S. Rosenberg. 2002. A balanced view of scale in

spatial statistical analysis. Ecography 25: 626-640.

Environmental Systems Research Institute, Inc. 2007. ArcGIS 9.2 [software]. ESRI,

Redlands, California.

Fancy, S. G. 1980. Nest tree selection by red squirrels in a boreal forest. Canadian Field

Naturalist 94: 198.

Fauth, P. T., D. G. Krementz, and J. E. Hines. 1991. Ectoparasitism and the role of green

nesting material in the European starling. Oecologia 88: 22-29.

Fisher, J. T., S. Boutin, and S. J. Hannon. 2005. The protean relationship between boreal

forest structure and red squirrel distribution at multiple spatial scales. Landscape

Ecology 20: 73-82

Fitzgibbon, C. D. 1993. The distribution of grey squirrel dreys in farm woodland: the

influence of wood area, isolation and management. Journal of Applied Ecology 30:

763-742.

Flyger, V., and J. E. Gates. 1982. Pine squirrels Tamiasciurus hudsonicus and T. douglasii.

Pp. 230-238 in Wild Mammals of North America: Biology, Management, and 85

Economics (J. A. Chapman and G. A. Feldhamer, eds.). The Johns Hopkins

University Press, Baltimore.

Fokidis, H. B., and T. S. Risch. 2005. The use of nest boxes to sample arboreal vertebrates.

Southeastern Naturalist 4: 447-458.

Forman, R. T. T. 1995. Land Mosaics: The Ecology of Landscapes and Regions. Cambridge

University Press, Cambridge, UK.

Friesen, L., M. D. Cadman, R. J. MacKay. 1999. Nesting success of neotropical migrant

songbirds in a highly fragmented landscape. Conservation Biology 13: 338-346.

Giez, I., O. L. Lange, and P. Proksch. 1994. Growth retarding activity of lichen substances

against the polyphagous herbivorous insect Spodoptera littoralis. Biochemical

Systematics and Ecology 22: 113-120.

Goheen, J. R., R. K. Swihart, and J. Robins. 2003a. Anatomy of a range expansion:

latitudinal changes in foraging behavior and cranial morphology of North American

red squirrels. Oikos 102: 3-44.

Goheen, J. R., R. K. Swihart, T. M. Gehring, and S. Miller. 20036. Forces structuring tree

squirrel communities in landscapes fragmented by agriculture: species differences in

perceptions of forest connectivity and carrying capacity. Oikos 102: 95-103.

Goheen, J. R., and R. K. Swihart. 2005. Resource selection and predation of North American

red squirrels in deciduous forest fragments. Journal of Mammalogy 86: 22-28.

Goodwin, B. J., D. J. Bender, T. A. Contreras, L. Fahrig, and J. F. Wegner. 1999. Testing for

habitat detection distances using orientation data. Oikos 84: 160-163.

Greenwood, P. J. 1980. Mating systems, philopatry, and dispersal in birds and mammals.

Animal Behaviour 28: 1140-1162. 86

Gu, W., and R. K. Swihart. 2004. Absent or undetected? Effects of non-detection of species

occurrence on wildlife-habitat models. Biological Conservation 116: 195-203.

Gustafson, E. J., and G. R. Parker. 1992. Relationships between landcover proportion and

indices of landscape spatial pattern. Landscape Ecology 7: 101-110.

Gustafson, E. J., and G. R. Parker. 1994. Using an index of habitat patch proximity for

landscape design. Landscape and Urban Planning 29: 117-130.

Hackett, H. M., and J. F. Pagels. 2003. Nest site characteristics of the endangered northern

flying squirrel (Glaucomys sabrinus coloratus) in southwest Virginia. American

Midland Naturalist 150: 321-331.

Haddad, N. M. 1999. Corridor use predicted from behaviors at habitat boundaries. American

Midland Naturalist 153: 215-227.

Hale, M. L., P. W. W. Lurz, M. D. F. Shirley, S. Rushton, R. M. Fuller, and K. Wolff. 2001.

Impact of landscape management on the genetic structure of red squirrel populations.

Science 293: 2246-2248.

Hanski, I. 1994. A practical model of metapopulation dynamics. The Journal of Animal

Ecology 63: 151-162.

Hargis, C. D., J. A. Bissonette, and J. L. David. 1998. The behavior of landscape metrics

commonly used in the study of habitat fragmentation. Landscape Ecology 13:

167-186.

Harris, L. D. 1984. The Fragmented Forest: Island Biogeography Theory and The

Preservation of Biotic Diversity. University of Chicago Press, Chicago, Illinois, USA.

Haughland, D. L., and K. W. Larsen. 2004a. Ecology of North American red squirrels across

contrasting habitats: relating natal dispersal to habitat. Journal of Mammalogy 85: 87

225-236.

Haughland, D. L., and K. W. Larsen. 20046. Exploration correlates with settlement: red

squirrel dispersal in contrasting habitats. Journal of Animal Ecology 73: 1024-1034.

Havelka, M. A., and J. S. Millar. 1996. Multiple nest boxes per station in the study of

Peromyscus leucopus. American Midland Naturalist 135: 376-380.

Hay ward, G. D., and R. Rosentreter. 1994. Lichens as nesting material for northern flying

squirrels in the northern Rocky Mountains. Journal of Mammalogy 75: 663-673.

Hemmes, R. B., A. Alvarado, and B. L. Hart. 2002. Use of California bay foliage by wood

rats for possible fumigation of nest-borne ectoparasites. Behavioural Ecology 13:

381-385.

Hilton, G. M., M. H. Hansell, G. D. Ruxton, J. M. Reid, and P. Monaghan. 2004. Using

artificial nests to test importance of nesting material and nest shelter for incubation

energetics. The Auk 121: 777-787.

Holland, J. D., D. G. Bert, and L. Fahrig. 2004. Determining the spatial scale of species

response to habitat. BioScience 54: 227-233.

Holloway, G. L. 2006. Flying squirrel (Glaucomys sabrinus and G. volans) habitat use and

ecology in landscapes managed with partial harvesting silviculture in central Ontario.

PhD dissertation, University of Toronto, Toronto, Ontario, Canada. Pp. 145.

Holloway, G. L., and J. R. Malcolm. 2006. Sciurid habitat relationships in forests managed

under selection and shelterwood silviculture in Ontario. Journal of Wildlife

Management 70: 1735-1745.

Holloway, G. L., and J. R. Malcolm. 2007. Nest-tree use by northern and southern flying

squirrels in central Ontario. Journal of Mammalogy 88: 226-233. 88

Holt, R. F., and K. Martin. 1997. Landscape modification and patch selection: the

demography of two secondary cavity nesters colonizing clearcuts. Auk 114: 443-455.

Hosmer, D.W., and S. Lemeshow. 2000. Applied Logistic Regression, 2nd ed. John Wiley

and Sons, New York, New York.

Imbeau, L., M. Monkkonen, and A. Desrochers. 2001. Long-term effects of forestry on birds

of the eastern Canadian boreal forests: a comparison with Fennoscandia.

Conservation Biology 15: 1151-1162.

Ingolfsdottir, K., G. A. Chung, V. G. Skulason, S. R. Gissurarson, and M. Vilhelmsdottir.

1998. Antimycobacterial activity of lichen metabolites in vitro. European Journal of

Pharmaceutical Science 6: 141-144.

Keita, S. M., C. Vincent, J. R Schmidt, and J. T. Arnason. 2001. Insecticidal effects of Thuja

occidentalis (Cupressaceae) essential oil on Callosobruchus maculatus [Coleoptera:

Bruchidae]. Canadian Journal of Plant Science 81: 173-177.

Kemp, G. A., and L. B. Keith. 1970. Dynamics and regulation of red squirrel {Tamiasciurus

hudsonicus) populations. Fxology 51: 763-779.

Klugh, A. B. 1927. Ecology of the red squirrel. Journal of Mammalogy 8: 1-32.

Krebs, C. J., B. L. Keller, and R. H. Tamarin. 1969. Microtus population biology:

demographic changes in fluctuating populations of Microtus ochrogaster and M.

pennsylvanicus in southern Indiana. Ecology 50: 587-607.

Lafuma, L., M. M. Lambrechts, and M. Raymond. 2001. Aromatic plants in bird nests as a

protection against blood-sucking flying insects? Behavioural Processes 56: 113-120.

Larsen, K.W., and S. Boutin. 1994. Movements, survival, and settlement of red squirrel

{Tamiasciurus hudsonicus) offspring. Ecology 75: 214-223. 89

Larsen, K.W., and S. Boutin. 1995. Exploring territory quality in the North American red

squirrel through removal experiments. Canadian Journal of Zoology 73: 1115-1122.

Laurance, W. F. 1991. Ecological correlates of extinction proneness in Australian tropical

rainforest mammals. Conservation Biology 5: 79-89.

Laurance, W. F. 1995. Extinction and survival of rainforest mammals in a fragmented tropical

landscape. Pp. 46-63 in Landscape Approaches in Mammalian Ecology and

Conservation (W. Z. Lidicker, Jr, ed.). University of Minnesota Press, Minnesota,

USA.

Lauterwein, M., M. Oethinger, K. Belsner, T. Peters, and R. Marre. 1995. In vitro activities

of the lichen secondary metabolites vulpinic acid, (C)- usnic acid, and (K)-usnic acid

against aerobic and anaerobic microorganisms. Antimicrobial Agents and

Chemotherapy 39: 2541-2543.

Lavers, A. L. 2004. Spatial ecology in a northern disjunct population of southern flying

squirrel, Glaucomys volans. MSc thesis, Acadia University, Wolfville, Nova Scotia,

Canada. Pp. 131.

Layne, J. N., and M. A. V. Raymond. 1994. Communal nesting of southern flying squirrels in

Florida. Journal of Mammalogy 75: 110-120.

Layne, J. N. 1954. The biology of the red squirrel, Tamiasciurus hudsonicus loquax (Bangs),

in central New York. Ecological Monographs 24: 227-268.

Levins, R. 1969. Some demographic and genetic consequences of environmental

heterogeneity for biological control. Bulletin of the Entomological Society of

America 15: 237-240

Levins, R. 1970. Extinction. Pp. 77-107 in Lectures on Mathematics in the Life Sciences (M. Gerstenhaber, ed.). American Mathematics Society, Providence, Rhode Island, USA.

Loeb, S. C, F. H. Tainter, and E. Cazares. 2000. Habitat associations of hypogeous fungi in

the southern Appalachians: implications for the endangered northern flying squirrel

(Glaucomys sabrinus coloratus). American Midland Naturalist 144: 286-296.

Lomolino, M. V. 1990. The target area hypothesis: the influence of island area immigration

rates of non-volant mammals. Oikos 57: 297-300.

MacArthur, R. H., and E. O. Wilson. 1967. The Theory of Island Biogeography. Princeton

University Press, Princeton, New Jersey, USA.

MacKenzie, D. I. 2005. What are the issues with presence-absence data for wildlife

managers? Journal of Wildlife Management 69: 849-860.

Male, S. K., J. Jones, and R. J. Robertson. 2006. Effects of nest-box density on the behavior

of tree swallows during nest building. Journal of Field Ornithology 77: 61-66.

Mannan, R. W, E. C. Meslow, and H. M. Wight. 1980. Use of snags by birds in Douglas-fir

forests, western Oregon. Journal of Wildlife Management 44: 787-797.

Maser, C, G. Anderson, and K. Cromack, Jr. 1979. Dead and downy woody material. Pp.

78-95 in Wildlife Habitats in Managed Forests: The Blue Mountains of Oregon and

Washington. U.S. Department of Agriculture, Forest Service Agriculture Handbook,

No. 553.

Maser, C, R. Anderson, and E. F. Bull. 1981. Aggregation and sex segregation in northern

flying squirrels in northeastern Oregon, an observation. Murrelet 62: 54-55.

McComb, W, and D. Lindenmayer. 1999. Dying, dead, and down trees. Pp. 335-375 in

Maintaining Biodiversity in Forest Ecosystems (M. L. Hunter, ed.). Cambridge

University Press, Cambridge, UK. 91

McEachen, T. J. 2007. A genetic test of range expansion by the southern flying squirrel

(Glaucomys volans) at its northern range boundary. M. Sc. thesis, Trent University,

Peterborough, Ontario, Canada. Pp. 46.

McGarigal, K., and B. J. Marks. 1995. FRAGSTATS: spatial pattern analysis program for

quantifying landscape structure. USDA Forest Service General Technical Report

PNW-351. Pp. 122.

Mortberg, U. 2001. Resident bird species in urban forest remnants; landscape and habitat

perspectives. Landscape Ecology 16: 193-203.

Mowrey, R. A., and J. C. Zasada. 1984. Den tree use and movements of northern flying

squirrels in interior Alaska and implications for forest management. Pp. 351-356 in

Fish and Wildlife Relationships in Old-growth Forests: Proceedings of a Symposium

(W. R. Meehan, T. R. Merrell, and T. A. Hanley, eds.). American Institute of Fisheries

Research Biology, Moorhead City, North Carolina.

Murrell, B. P., L. A. Durden, and J. A. Cook. 2003. Host associations of the tick, Ixodes

angustus (Acari: Ixodidae), on Alaskan mammals. Journal of Medical Entomology

40: 682-685.

Muul, I. 1974. Geographic variation in the nesting habits of Glaucomys volans. Journal of

Mammalogy 55: 840-844.

Newton, I. 1994. Experiments on the limitation of bird breeding densities: a review. Ibis 136:

397-411.

Nishtoba, Y., H. Nishmura, T. Nishiyama, and J. Mizutani. 1987. Lichen acids, plant growth

inhibitors from Usnea longissima. Phytochemistry 26: 3181-3185.

Nupp, T. E., and R. K. Swihart. 2000. Landscape-level correlates of small mammal assemblages in forest fragments of farmland. Journal of Mammalogy 81: 512-526.

O'Connell, A. R, Jr., F. A. Servello, J. Higgins, and W. Halteman. 2001. Status and habitat

relationships of northern flying squirrels on Mount Desert Island, Maine.

Northeastern Naturalist 8: 127-136.

Okuyama, E., K. Umeyama, M. Yamazaki, Y. Kinoshita, and Y. Yamamoto. 1995. Usnic acid

and diffractaic acid as analgesic and antipyretic components of Usnea diffracta.

Planta Medica61: 113-115.

Paton, P. 1994. The effect of edge on avian nest success: how strong is the evidence?

Conservation Biology 8: 17-26.

Pearce, C. M. 1992. Pattern analysis of forest cover in southwestern Ontario. East Lakes

Geographer 27: 65-76.

Pedlar, J. H., L. Fahrig, and H. G. Merriam. 1997. Raccoon habitat use at two spatial scales.

Journal of Wildlife Management 61: 102-112.

Pretzlaw, T., C. Trudeau, M. M. Humphries, J. M. LaMontagne, and S. Boutin. 2006. Red

squirrel (Tamiasciurus hudsonicus) feeding on spruce bark beetles (Dendroctonus

rufipennis): energetic and ecological implications. Journal of Mammalogy 87:

909-914.

Pung, O. J., L. D. Carlile, J. Whitlock, S. P. Vives, L. A. Durden, and E. Spadgenske. 2000.

Survey and host fitness effects of red-cockaded woodpecker blood parasites and nest

cavity arthropods. Journal of Parasitology 86: 506-510.

R Development Core Team. 2007. R: A Language and Environment for Statistical Computing

(version 2.5.1). R Foundation for Statistical Computing, Vienna, Austria.

Ransome, D. B. and T. P. Sullivan. 1997. Food limitation and habitat preference of 93

Glaucomys sabrinus and Tamiasciurus hudsonicus. Journal of Mammology 78:

538-549.

Riley, J. L., and P. Mohr. 1994. The natural heritage of southern Ontario's settled landscapes.

Technical Report TR-001. Science and Technology Transfer, Ontario Ministry of

Natural Resources, Aurora, Ontario. Pp. 78.

Rodgers, J. A., Jr., A. S. Wenner, and S. T. Schwikert. 1988. The use and function of green

nest material by wood storks. Wilson Bulletin 100: 411-423.

Rosenberg, D. K., J. D. Fraser, and D. F. Stauffer. 1988. Use and characteristics of snags in

young and old forest stands in southwest Virginia. Forest Science 34: 224-228.

Rosenberg, D. K., and R. G. Anthony. 1992. Characteristics of northern flying squirrel

populations in young second- and old-growth forests in western Oregon. Canadian

Journal of Zoology 70: 161-166.

Rothwell, R. 1979. Nest sites of red squirrels (Tamiasciurus hudsonicus) in the Laramie

Range of southeastern Wyoming. Journal of Mammalogy 60: 404-405.

Rowe, J. S. 1972. Forest Regions of Canada. Department of the Environment, Canadian

Forest Service. Publication Number 1300. Pp. 172.

Rusch, D. A., and W. G. Reeder. 1978. Population ecology of Alberta red squirrels. Ecology

59: 400-420.

Rust, H. J. 1946. Mammals of northern Idaho. Journal of Mammalogy 27: 308-327.

SAS Institute Inc. 2000. Statistical Analysis System (SAS) User's Guide: Statistics (version

8). SAS Institute Inc., Cary, North Carolina.

Selonen, V., I. K. Hanski, and P. C. Stevens. 2001. Space use of the Siberian flying squirrel

Pteromys volans in fragmented forest landscapes. Ecography 24: 588-600. Selonen, V., and I. K. Hanski. 2003. Movements of the flying squirrel Pteromys volans in

corridors and in matrix habitat. Ecography 26: 641-651.

Selonen, V., and I. K. Hanski. 2004. Young flying squirrels {Pteromys volans) dispersing in

fragmented forests. Behavioral Ecology 15: 564-571.

Selonen, V., and I. K. Hanski. 2006. Habitat exploration and use in dispersing juvenile flying

squirrels. Journal of Applied Ecology 75: 1440-1449.

Shaw, A. C. 1953. The essential oil of Thuja occidentalis. Canadian Journal of Chemistry 31:

277-281.

Sheperd, B. E, and R. K. Swihart. 1995. Spatial dynamics of fox squirrels (Sciurus niger) in

fragmented landscapes. Canadian Journal of Zoology 73: 2098-2105.

Smith, M. C. 1968. Red squirrel responses to spruce cone failure in interior Alaska. Journal

of Wildlife Management 32: 305-316.

Smith, W. R, S. M. Gende, and J. V. Nichols. 2004. Ecological correlates of flying squirrel

microhabitat use and density in temperate rainforests of southeastern Alaska. Journal

of Mammalogy 85: 663-674.

Smith, W. R, and D. K. Person. 2007. Estimated persistence of northern flying squirrel

populations in temperate rain forest fragments of southeast Alaska. Biological

Conservation 137: 626-636.

Sonenshine, D. E., D. G. Cerretani, G. Enlow, and B. L. Elisberg. 1973. Improved methods

for capturing wild flying squirrels. Journal of Wildlife Management 37: 588-590.

Soule, M. E. 1980. Thresholds for survival: maintaining fitness and evolutionary potential.

Pp. 153-169 in Dynamics of Extinction (E. K. Elliot, ed.). John Wiley and Sons, New

York, New York, USA. 95

Stamps, J. A., M. Buechner, and V. V. Krishnan. 1987. The effects of edge permeability and

habitat geometry on emigration from patches of habitat. American Naturalist 129:

533-552.

Stapp, P., P. J. Pekins, and W. W. Mautz. 1991. Winter energy expenditure and the

distribution of southern flying squirrels. Canadian Journal of Zoology 69: 2548-2555.

Sullivan, T. P. and R. A. Moses. 1986. Red squirrel populations in natural and managed

stands of lodgepole pine. Journal of Wildlife Management 50: 595-601.

Sun, C. 1997. Dispersal of young in red squirrels {Tamiasciurus hudsonicus). American

Midland Naturalist 138: 252-259.

Sutherland, G. D., A. S. Harestad, K. Price, and K. P. Lertzman. 2000. Scaling of natal

dispersal distances in terrestrial birds and mammals. Conservation Ecology 4: 16.

[online] URL: http://www.ecologyandsociety.org/vol4/issl/artl6/.

Swihart, R. K., andT. E. Nupp. 1998. Modeling population responses of North American tree

squirrels to agriculturally induced fragmentation of forests. Pp. 1-20 in Ecology and

Evolutionary Biology of Tree Squirrels (M. A. Steele, J. F. Merritt, and D. A. Zegers,

eds.). Virginia Museum of Natural History Special Publication 6: 1-310.

Swihart, R. K., J. J. Lusk, J. E. Duchamp, C. E. Rizkalla, and J. E. Moore. 2006. The roles of

landscape context, niche breadth, and range boundaries in predicting species

responses to habitat alteration. Diversity and Distributions 12: 277-287.

Swihart, R. K., J. R. Goheen, S. A. Schnelker, and C. E. Rizkalla. 2007. Testing the

generality of patch and landscape-level predictors of tree squirrel occurrence at a

regional scale. Journal of Mammalogy 88: 564-572.

Taulman, J. F. 1999. Selection of nest trees by southern flying squirrels (Sciuridae: 96

Glaucomys volans) in Arkansas. Journal of Zoology 248: 369-377.

Tay, T., A. O. Turk, M. Yilmaz, H. Turk, and M. Kivanc. 2004. Evaluation of the

antimicrobial activity of the acetone extract of the lichen Ramalina farinacea and its

(+)usnic acid, norstictic acid, and protocetraric acid constituents. Zeitschrift Fur

Naturforschung C-A Journal of Biosciences 59: 384-388.

Ter Braak, C. J. R, and P. Smilauer. 1998. Canoco Reference Manual and User's Guide to

Canoco for Windows: Software for Canonical Community Ordination (version 4).

Microcomputer Power, Ithaca, New York, USA.

Vahle, J. R., and D. R. Patton. 1983. Red squirrel cover requirements in Arizona mixed

coniferous forests. Journal of Forestry 81:14-15. van Apeldoorn, R. C, X. Celada, and W. Nieuwenhuizen. 1994. Distribution and dynamics of

red squirrel (Sciurus vulgaris L.) in a landscape with fragmented habitat. Landscape

Ecology 9: 227-235.

Villa, L. J., A. B. Carey, T. M. Wilson, and K. E. Glos. 1999. Maturation and reproduction of

northern flying squirrels in Pacific Northwest forests. United States Department of

Agriculture, Forest Service, General Technical Report PNW-GTR-444. Pp. 59.

Vos, C. C, J. Verboom, P. F. M. Opdam, and C. J. F. Ter Braak. 2001. Toward ecologically

scaled landscape indices. The American Naturalist 183: 24-41.

Waters, J. R., and C. J. Zabel. 1995. Northern flying squirrel densities in fir forests of

northeastern California. Journal of Wildlife Management 59: 858-866.

Wauters, L., P. Casale, and A. A. Dhondt. 1994. Space use and dispersal of red squirrels in

fragmented habitat. Oikos 69: 140-146.

Weigl, P. D., and D. W. Osgood. 1974. Study of the northern flying squirrel Glaucomys 97

sabrinus by temperature telemetry. American Midland Naturalist 92: 482-486.

Weigl, P. D. 1978. Resource overlap, interspecific interactions and the distribution of the

flying squirrels, Glaucomys volans and G. sabrinus. American Midland Naturalist

100: 83-96.

Weigl, P. D. 2007. The northern flying squirrel {Glaucomys sabrinus): a conservation

challenge. Journal of Mammalogy 88: 897-907.

Wells-Gosling, N., and L. R. Heaney. 1984. Glaucomys sabrinus. Mammalian Species 229:

1-8.

Wheatley, M., J. T. Fisher, K. W Larsen, J. Litkes, and S. Boutin. 2005. Using GIS to relate

small mammal abundance and landscape structure at multiple spatial extents: the

northern flying squirrel in Alberta, Canada. Journal of Applied Ecology 42: 577-586.

Whitcomb, R. R, J. F. Lynch, M. K. Klimkiewicz, C. S. Robbins, B. L. Whitcomb, and D.

Bystrak. 1981. Effects of forest fragmentation on avifauna of the eastern deciduous

forest. Pp. 125-206 in Forest Island Dynamics in Man-dominated Landscapes (R. L.

Burgess and D. M. Sharpe, eds.). Springer-Verlag, New York, New York, USA.

Wiebe, K. L. 2001. Microclimate of tree cavity nests: is it important for reproductive success

in northern flickers? Auk 118: 412-421.

Wiens, J. A., C. S. Crawford, and J. R. Gosz. 1985. Boundary dynamics: a conceptual

framework for studying landscape ecology. Oikos 45: 421-427.

Wiens, J. A., and B. T. Milne. 1989. Scaling of landscapes in landscape ecology, or landscape

ecology from a beetle's perspective. Landscape Ecology 3: 87-96.

Wiens, J. A., N. C. Stenseth, B. Van Home, and R. A. Ims. 1993. Ecological mechanisms and

landscape ecology. Oikos 66: 369-380. 98

Wiens, J. A. 1996. Wildlife in patchy environments: metapopulations, mosaics, and

management. Pp. 53-84 in Metapopulations and Wildlife Conservation (D. R.

McCullough, ed.). Island Press, Washington, DC, USA.

Wimberger, P. H. 1984. The use of green plant material in bird nests to avoid ectoparasites.

Auk 101: 615-618.

Witt, J. W. 1992. Home range and density estimates for the northern flying squirrel,

Glaucomys sabrinus, in western Oregon. Journal of Mammalogy 73: 921-929.

Witte, L., J. Berlin, V. Wray, W Schubert, W. Kohl, G. Hone, and J. Hammer. 1983. Mono-

and diterpenes from cell cultures of Thuja occidentalis. Planta Medica 49: 216-221.

Wolff, J. O. 1999. Behavioral model systems. Pp. 11-40 in Landscape Ecology of Small

Mammals (G. W. Barrett and J. D. Peles, eds.). Springer-Verlag, New York, New

York, USA.

Yatagai, M., T. Sato, and T. Takahashi. 1985. Terpenes of oils from Cupressaceae.

Biochemical Systematics and Ecology 13: 377-385.

Yilmaz, M., A. O. Turk, T. Tay, and M. Kivanc. 2004. The antimicrobial activity of extracts

of the lichen Cladonia foliacea and its (-)usnic acid, atranorin, and fumarprotocetraric

acid constituents. Zeitschrift Fur Naturforschung C-A Journal of Biosciences 59:

249-254.