WITHIN PEATLAND SPATIAL STRUCTURING AND THE INFLUENCE OF THE MATRIX ON BETWEEN PEATLAND MOVEMENT OF THE DRAGONFLY, Leucorrhinia hudsonica IN WESTERN NEWFOUNDLAND

by

KRISTA S. CHIN B.Sc. McGill University, 1999

Thesis submitted in partial fulfillment of the requirements for the Degree of Masters of Science (Biology)

Acadia University Spring Convocation 2006

© by KRISTA S. CHIN, 2006 ii

This thesis by KRISTA S. CHIN was defended successfully in an oral examination on December 16, 2005.

The examining committee for the thesis was:

Dr. John Roff, Chair

Dr. Marina Silva, External Reader

Dr. Sara Good-Avila, Internal Reader

Dr. Philip Taylor, Supervisor

Dr. Dave Shutler, Acting Head

This thesis is accepted in its present form by the Division of Research and Graduate Studies as satisfying the thesis requirements for the degree of Master of Science (Biology).

iii

Table of Contents

List of Figures...... iv

List of Tables ...... vi

Abstract...... vii

Acknowledgements...... vii

General Introduction...... 1

Chapter 1. Spatial structuring and net movement distances of Leucorrhinia hudsonica within peatlands in western Newfoundland...... 10

Abstract...... 10 Introduction...... 11 Methods...... 14 Results...... 19 Discussion...... 23

Chapter 2. Effects of three matrix types on the connectivity between peatlands for the peatland dragonfly, Leucorrhinia hudsonica, in western Newfoundland...... 41

Abstract...... 41 Introduction...... 42 Methods...... 46 Results...... 50 Discussion...... 52

General Discussion...... 67

References...... 69

Appendix 1 ...... 81 Appendix 2 ...... 82 Appendix 3 ...... 83

iv

List of Figures

General Introduction

Figure I.1. Map of Newfoundland indicating general study region. …………………..... 9

Chapter 1.

Figure 1.1. Aerial photograph of the 3 peatlands in the study area. ………………….. 30

Figure 1.2. Schematic diagram of peatland B divided into 32×32 m grids. ………….. 31

Figure 1.3. Semivariograms for peatlands B and M. …………………………………. 32

Figure 1.4. Deviations from the expected distribution of occupied pools in peatlands B and M. ………….………………………………………………………………. 33

Figure 1.5. Comparison of the deviations from the expected distribution of small and large pools to the observed distribution of all occupied pools in peatlands B and M. ………………………………………………………………………….. 34

Figure 1.6. Influence of pool area on the mean number of individuals found on pools in peatlands B and M. …………………………………………………..... 35

Figure 1.7. A. Comparison of θ across 7 spatial scales in the null models, B. Comparing two different methods of calculating θ, and C. Comparing θ between null models and models of best fit for peatlands B and M. …………... 36

Figure 1.8. Quantile-quantile plots comparing the net movement distances to the expected distribution of movement with 95% confidence envelope in peatlands B and M. …………………………………………………………….. 37

Chapter 2

Figure 2.1. Aerial photograph of the study area. ……………………………………... 59

Figure 2.2. The influence of distance on the proportion of animals moved. …………. 60

Figure 2.3. Interaction plot of the proportion of male L. hudsonica that moved between peatlands in 3 distance classes, according to the amount of FOREST MATRIX. …………………………………………………………………………. 61

Appendix

Appendix 1. Frequency of pool sizes in peatlands B, K, and M. ……………………... 81

v

Appendix 2. The relationship between the proportion of large pools in each grid to the total percent cover of water large pools accounted for at the 64×64 m2 scale for peatlands B, K, and M. ……………………………………………….. 82

vi

List of Tables

Chapter 1

Table 1.1. Physical characteristics and mark-release-recapture results for peatlands B, K, and M. ……………………………………………………………………. 38

Table 1.2. Summary of pool characteristics for peatlands B, K, and M. .……………. 38

Table 1.3. Parameter estimates (± SE) for the glm.nb and glm models of best fit. …... 39

Table 1.4. Residual deviances of the glm.nb and the glm models of best fit. ………… 40

Chapter 2

Table 2.1. Summary of the physical characteristics (± SE) and mark-release- recapture results for the north, south, and central peatlands. …………………... 62

Table 2.2. Daily survival, recapture, and movement rates (± SE) for the functional groups of peatlands. ……………………………………………………………. 63

Table 2.3. Top 4 MARK models ranked using QAICc. ………………………………. 64

Table 2.4. Parameter estimates (± SE) for the co-variates of the top 4 models of best fit. ………………………………………………………………………….. 65

Table 2.5. Parameter estimates (± SE) for the glm exploring the influence of DISTANCE and FOREST MATRIX on the number of male L. hudsonica moving between peatlands. ……………………………………………………………... 66

Table 2.6. Residual deviances for the glm exploring the influence of DISTANCE and FOREST MATRIX on the number of male L. hudsonica moving between peatlands. ………………………………………………………………………. 66

Appendix

Appendix 3. Counts of pools and of male L. hudsonica found on small and large pools within and beyond the mean pool to peatland boundary distance. …….. 83

vii

Abstract

It is important to quantify how species respond to landscape change. I examined the influence of fine and broad scale structures on the distribution and movements of the peatland dragonfly, Leucorrhinia hudsonica in Newfoundland using mark-recapture techniques.

Within peatlands there was a positive, non-linear relationship between pool size and number of male L. hudsonica. Results suggested that net movement distances were random at small scales, but less than expected at larger scales. Large pools are preferred oviposition sites, and were the focus of male territorial activity, though most territories encompassed a number of adjacent pools.

At the landscape scale, there was more emigration out of peatlands having less water.

Though ambiguous, model selection indicated that forest impeded movement more than clearcut or scrub matrix, and that scrub facilitated movement. There was an interaction between matrix type and distance between peatlands; indicating that movement of males through the matrix may not be a simple function of habitat permeability. Rather, it may be influenced by the different behaviours affecting decisions at different spatial scales.

viii

Acknowledgements

First, I would sincerely like to thank Leo and Eliza Hynes for their generous nature as well as letting us set up camp in (and all around) their cabin on Island Pond.

Thanks to Stephen Flemming and Irene Howell (Gros Morne National Park), Adele Mullie (Acadia), Faron Knott (Corner Brook Pulp and Paper), Greg Robertson (CWS), and Ian Warkentin (Sir Wilfred Grenfell College), who provided logistical support, advice, and/or assistance to the project.

Many many thanks to the Powells for letting us stay at their home during our “days off”, feeding us, and especially for helping us bring in and bring out camp and trusting me to dog-sit their hearing impaired and behaviourally challenged dog, Casey.

The field season was a success because of my ever positive and sometimes funny field assistants. Marla MacAulay and Donald Humphrey kept me on my toes all season. (“I’m watching you guys.”) Much help was provided by Ray Reid, Scott Howell, Chris Shears, Shelley Garland, Dr. Becky Achtman, Nicole Davis, Ryan Sharpe, Meagan Rivers, James Loughlin, Ed MacAulay, Sarah MacAulay, and Sheldon Pittman.

Special thanks to Michelle McPherson (hide the cookies!), Darroch Whitaker, Joe Nocera, Anna Calvert, and Tom Herman who helped with the development of ideas and/or statistical problems. I am indebted to Ian Jonsen who clarified many theoretical concepts and gave tonnes of statistical advice.

A big acknowledgement to the unique group of grad students whose friendship and wit kept me laughing for the extended period of time I was at Acadia; Katie Dalley, Tina Leonard, Greg Mitchell, Jenny Berlo, Kim “the flasher” Dawe, Shannon “Wimbledon anyone?” O’Connor, Jen Lusk, Brennan Caverhill, Sean LeMoine, Dragica Matkovich, Jeff Reader, and Janine Beckett. A special nod to the “theoretical Class of 2004”; Kristin “Baby K” Powell, who taught me about theory, Mike “I don’t do potlucks” Peckford who was (to everyone’s surprise) right about the definition of “inflammable”, and Andrew “The Gock- soft ‘g’” Trant whose perfect bow will surely impress his future in-laws.

I would also like to thank the Acadia Women’s Soccer Team for 2 years of great fun, competition, and friendship. In the words of Farqy: “This team is full of unmatched character… we are in fact, everything a team is, should, could, and ideally ever will be.”

A million thanks to Mom, Dad, Nat, Karen, and Marie for support during my thesis, but mostly for the care packages you sent (i.e. life support system) while I was in the Main.

Lastly, thanks to Phil Taylor for his ideas, guidance, and giving me many opportunities to learn from this project while not letting me get “bogged down with the stats”.

~ Il faut avoir des mauvaises journées pour avoir des bonnes journées. ~ Today is a good day.

1

General Introduction

Empirical studies are not only necessary to test and improve theoretical models, but are essential for increasing our understanding of how natural populations respond to human- induced habitat degradation, destruction, and fragmentation at multiple temporal and spatial scales. With clear questions and a robust study design, such research has the power to influence management (Jonsen et al. 2001, Purse et al. 2003).

Landscape heterogeneity and spatial scales

Landscapes are heterogeneous areas that exist over a range of spatial scales and are made up of discrete patches nested within each other (Senft et al. 1987, Turner 1989). It is important to note that patches are scale dependent (Kotilar and Wiens 1990). That is, they are commonly identified at a spatial scale that is relevant to the species, even if the resource itself does not appear to be patchy at finer scales. For example, male dragonflies aggregate around pools in peatlands to breed (Corbet 1999). Within a peatland, pools may be randomly distributed. However, pools may appear patchily distributed at broader spatial scales because peatlands are heterogeneously distributed throughout the landscape.

There is no correct single scale at which to study the ecology of species (Turner 1989,

Wiens 1989, Levin 1992). Every organism interacts with the environment at its own unique set of scales, so multiple ecological scales that are independently related to different sets of ecological processes should be inspected (Addicott et al. 1987, Turner et al. 1989, Holling 1992, Kareiva and Wennergren 1995, Allen and Holling 2002, Kadoya

2 et al. 2004). Furthermore, information from small scales can not necessarily be obtained in broader-scale experiments or surveys, nor can one extrapolate from small-scale experiments or surveys to broader scales and vice versa (Wiens et al. 1993).

At all scales, the behaviours and spatial distribution that populations exhibit are a function of the biology of the organism and structure of the landscape (Wiens et al.

1997). At small spatial scales, the distributions of animals are influenced by local processes (e.g. habitat selection, foraging, mating, territorial and oviposition behaviours, local movement; (Jonsen and Fahrig 1997, Conrad et al. 1999, Roslin 2000). At broader spatial scales, population-level processes (e.g. broad-scale movement, environmental effects, natural selection) influence the spatial distribution of animals (Grover and

Thompson 1986, Jonsen and Fahrig 1997, Schwarz et al. 2003). For example, krill distribution at fine scales is maintained by cues from conspecifics that promote schooling, which offers them protection against predators (Hamner and Hamner 2000).

However, at larger spatial scales, their spatial distribution can be explained by oceanic currents and ocean water temperature (Flierl et al. 1999).

Additionally, cross-scale impacts can alter other population level processes or structure.

For example, Senft et al. (1987) showed that large herbivores select landscape units that are rich in resources. Within those, they choose the most productive patch and, within those, the most palatable species. Each decision is at a particular scale and involves multiple processes. They decide where in the landscape they will forage, how far from

3 water they will travel, the steepness of slopes they will traverse, how close they will allow predators to approach, and which parts of which plants they will eat.

As an extension, because organisms exist in various body sizes, animals living in the same landscape perceive, exploit, and structure the environment at different scales

(Lindstedt et al. 1986, Holling 1992, With 1994, Jetz et al. 2004). Increasing our understanding of how organisms interact with and structure their environment at different scales is thus a fundamental aspect of spatial ecology.

Movement, connectivity, and matrix

Movement is recognized as a fundamental process, for all species, that operates over multiple spatial and temporal scales (Roslin 2000). It influences both the spatial structure and regional dynamics of a population (Senft et al. 1987, Krawchuk and Taylor 2003).

Multiple biotic and abiotic factors act upon individuals to influence their movement behaviours. These factors range from the individual’s need to eat (Ward and Saltz 1994), access habitat (Dunning et al. 1992, Cronin 2003), and avoid predation (McIvor and

Odum 1988), to escaping long-term environmental change (e.g. global climate change)

(Holt 1990).

Animal movement between two habitat patches may be facilitated (Pither and Taylor

1998, Roland et al. 2000) or constrained (Fahrig and Merriam 1985, Ricketts 2001, Purse et al. 2003) by the matrix type (habitat/degraded habitat) that separates two patches.

Because landscapes are heterogeneous and resources are patchily distributed in space,

4

animals moving from one habitat to another may encounter matrix types that offer

varying levels of hostility (Bélisle and St. Clair 2001, Jonsen et al. 2001, Jules and

Shahani 2003). Therefore, a patch that is separated by one matrix type may functionally be more isolated than another habitat patch of similar quality but separated by a different matrix type.

Recent studies have established that landscape connectivity, the ease in which a species moves through the landscape (Taylor et al. 1993), is affected by matrix type (Yang 2000,

Bélisle and St. Clair 2001, Jonsen et al. 2001). That is, there is a direct relationship between movement rates and matrix type, and matrix type can influence the effective isolation of a patch beyond that of distance alone (Hof and Flather 1996, Roland et al.

2000, Ricketts 2001).

Increased isolation due to the structure of the matrix may reduce the ability of individuals to disperse which may ultimately lead to decreased gene flow among populations

(Hastings and Harrison 1994, Bohonak 1999, Sork et al. 1999, Hale et al. 2001).

Conversely, if movement rates are high, populations may lose their unique subpopulation dynamics resulting in synchrony among populations (Lande et al. 1999). The higher the movement rate, the more homogenized growth rates tend to become and the more synchronous the dynamics (Earn et al. 2000). The main effect is that a regional population is more likely to go extinct when synchrony is high because when low population densities align with some stochastic event, there can be no rescue effect

(Courchamp et al. 1999, Earn et al. 2000).

5

There have been few empirical studies that have looked at the effect of matrix on odonate movement rates. Both Pither and Taylor (1998) and Jonsen and Taylor (2000b) performed manipulative experiments on the same two ecologically similar species of damselflies (Calopteryx spp.). Pither and Taylor (1998) transferred adults from their stream habitat across forest and pasture matrices. Connectivity was measured as the proportion of animals that returned to the stream. Jonsen and Taylor (2000b) also performed manipulative experiments to compare the movement behaviour across three landscape types (forested, partially forested, and non-forested habitats). Most recently,

Purse et al. (2003) indirectly explored the effects of matrix type on movement rates through a mark-release-recapture survey. They examined dispersal of a damselfly for which habitat patches were separated by patches of scrub, valley mire, and/or heath.

Collectively, these studies demonstrated that matrix type influenced movement rates, but also showed that similar species and different sexes of the same species were influenced dissimilarly by the quality of the matrix (Jonsen and Taylor 2000b, Jonsen and Taylor

2000a).

The Greater Gros Morne Ecosystem

The present study was conducted in the Humber River watershed, located east of Gros

Morne National Park (GMNP), in the Gros Morne Greater Ecosystem (GMGE)

(Universal Transverse Mercator 5501000, 0478000) of Newfoundland, Canada

(Figure I.1). Most of the forest east and south of the park is being actively harvested.

The area is a naturally patchy landscape consisting of gap-replacing old growth boreal

6 forest dominated by balsam fir (Abies balsamea) and black spruce (Picea mariana) interspersed with peatlands, lakes and streams, as well as clearcuts of various regeneration stages and sizes (McCarthy 2001). The area has been subjected to intense tree harvesting activities since 1991 and the long-term effects of this forestry practice in the area are unknown.

Peatlands in this system are dominated by Sphagnum mosses but have extensive areas of shrubs and many flowering plants (Holder 2001). They have distinct pools of water of varying sizes and depths scattered throughout. Dragonflies and damselflies use these areas as breeding habitats and larvae develop within the pools (Holder 2001).

Study Species

Leucorrhinia hudsonica is a small (ca. 3 cm long), univoltine, black-and-red (males) or black-and-yellow (females) dragonfly that is distributed throughout Canada and the central and northern part of the United States (Hilton 1987, Needham et al. 2000). The species is abundant throughout its range (Hilton 1987), including the GMGE (Holder

2001, McPherson 2003). These dragonflies inhabit marshes, boggy ponds and lakes, and peatland bogs and fens (Nikula et al. 2002). They are considered to be perchers; that is, they regulate their body temperature primarily by positioning their body at different angles to absorb or avoid solar radiation (Hilton 1984, Corbet 1999). Males are generally territorial, guarding areas of water throughout the day (Hilton 1984). Females oviposit in flight by taping the surface of the water with their abdomen (Hilton 1984).

7

Larvae live for approximately 2-3 years in peatland pools and go through 12 instars

before they emerge as tenerals (immature adults) (Larson and House 1990). Once the

tenerals’ wings have dried and stiffened, they make their maiden flight to the forest to

sexually mature (Corbet 1980). After maturation, males enter peatlands and concentrate

around oviposition sites (pools of water) for the majority of the day waiting for mating

opportunities (Hilton 1984, McPherson 2003). Females tend to be more cryptic. Once

mature, they spend most of their time in the forest foraging and roosting and will only

enter a peatland when they are ready to mate and oviposit (Hilton 1984). The lifespan of

adults ranges from approximately 2-4 weeks (Corbet 1999).

L. hudsonica are ideal animals for testing how movement rates are influenced by

landscape structure and how this affects population distribution. They are highly mobile

animals that are conspicuous, easy to handle, diurnal, and are required to move between

two habitat types during their adult life cycle; forest habitats to roost and forage and

peatland habitats for mating and for females to oviposit (Corbet 1999).

McPherson (2003) demonstrated that the matrix between nearest-neighbour peatlands

influenced the distribution of larval L. hudsonica beyond the effects of finer-scale

variables and distance between peatlands. There was a higher incidence and abundance

of larvae in peatlands that were separated by scrub than by forest, suggesting that forest

impeded movement more than scrub. Although McPherson (2003) was not able to model the influence of cut matrix on the incidence and abundance of L. hudsonica, the average proportion of occupied pools in peatlands that were separated by cut, fell between the

8

average proportion of occupied pools in peatlands that were separated by forest and scrub. This suggested that forest impeded movement more than cut and scrub, and that cut impeded movement more than scrub.

Because male and female L. hudsonica have different strategies to maximize their reproductive success, they are unevenly distributed within the landscape (Hilton 1984).

Males aggregate within peatlands and females spend extensive periods of time foraging in forests to increase their fecundity (Corbet 1999). These behavioural differences result in different movement rates between the sexes, and in a generally higher abundance of males being observed during the day. As a consequence, in this study I focused only on movement behaviour of males.

In chapter 1, I examined the spatial distribution of male L. hudsonica at the scale of the peatland using mark-release-recapture. I explored how physical characteristics of the pool as well as the processes of territorial behaviour and movement influenced distribution.

In chapter 2, I examined how the physical characteristics of the peatland and landscape, including matrix type (forest, harvested cutblocks, and scrub), influenced movement rates of male L. hudsonica between peatlands using mark-release-recapture.

9

N

150 km

Figure I.1. Map of the province of Newfoundland, Canada indicating the general location of the Humber River watershed (open black oval) (UTM 5501000, 0478000). The shaded area west of the study site represents Gros Morne National Park.

10

Chapter 1. Spatial structuring and net movement distances of Leucorrhinia hudsonica within peatlands in western Newfoundland.

Abstract

I examined how adult males of a peatland dragonfly, Leucorrhinia hudsonica, are

spatially distributed within peatlands and behavioural mechanisms (e.g. territorial

behaviour, movement) that potentially lead to the observed distribution at the fine scale.

A mark-release-recapture survey was carried out in three peatlands in western

Newfoundland. According to variance:mean ratios, populations in two of the three

peatlands were spatially autocorrelated. Ripley’s K suggested a positive relationship

between pool size and the number of individuals at the pool. Generalized linear models

with a negative binomial family showed decreasing aggregation of numbers of adults

with increasing spatial scale. The surface area of water consistently explained the

majority of the null deviance in all models. Lastly, quantile-quantile plots suggest that

net movement distances are random at small scales (between 0-60 m) and that there is

less movement than expected at larger scales. I propose that large pools (area > 5 m2), which act as preferred oviposition sites, are the main focus of territorial activity for males of this species, and their territoriality ranges over a number of pools generally found within the same area.

11

Introduction

Most animals move over multiple spatial and temporal scales (Ward and Saltz 1994, With

and Crist 1996, Jonsen and Taylor 2000a). Movement is a process that is fundamental to

the persistence of every population and is a major factor shaping the distribution of

animal populations in space and time (Harrison 1989, Krawchuk and Taylor 2003). An

individual’s need to find food, shelter, reproductive opportunities, minimize competition,

or avoid predation or unfavourable conditions will cause it to leave familiar surroundings

in search of other areas (Buchwald 1992, Browne et al. 1999, Joyce et al. 1999, Opdam

and Wascher 2004). Movement patterns differ among and within species and vary

according to physiological and behavioural states (McIntyre and Wiens 1999).

The degree to which movement appears to be structured is not only influenced by the

scale of observation (Turchin 1998), but also by the distribution of resources within the

landscape. Because resources are generally distributed non-homogenously within a

landscape (Lancaster et al. 2003, Dalthorp 2004), the distribution of animals requiring

specific resources will be influenced by their physical environment and their movement abilities at multiple spatial scales (Levin 1992, Roland and Taylor 1997, Jonsen and

Taylor 2000a, Krawchuk and Taylor 2003, Trzcinski et al. 2003). For example, dung beetles (genus Aphodius) require dung pats which are an ephemeral resource. Once a pat is depleted, beetles must move on in search of fresh pats. As a result, spatial structuring

of dung beetles not only exists on a temporal level (the duration of the dung pats), but on

two spatial levels; the individual dung pats in the pasture (fine-scale) and cattle pastures

across the landscape (broad-scale) (Roslin 2000). Thus the distribution of beetles is

12 affected by cattle behaviours operating at two scales, namely those behaviours affecting the timing and location of defecation within the pasture occupied by the herd, and those affecting the distribution of herds among pastures. Behavioural responses of beetles relative to the distribution of pats then mediate their distribution within the landscape.

Many studies have demonstrated that at the fine scale, for the species, along with the distribution of resources, differences in the abundance of suitable resources (e.g. food, water, shelter, breeding sites) influence the distribution of animals and plants (Colwell and Landrum 1993, Bean et al. 2002, Jepsen et al. 2002, Rietkerk et al. 2002, Dalthorp

2004). Areas with large amounts of resources have greater habitat quality which results in higher densities of organisms (Neubauer and Rehfeldt 1995, Kadoya et al. 2004).

Competition for preferred resources among odonates, particularly for breeding sites, generally results in territorial behaviour among males (Hilton 1984, Alcock 1987a).

Males that are able to successfully defend a breeding site will have more females visit their territories. Furthermore, males that are able to secure higher quality territories have an even higher success rate of passing on their genes (Alcock 1987a, b). For example,

Moore (1989) found that only one (1%) satellite male (i.e. non-territory holder) of

Libellula luctuosa successfully mated. Plainstow and Siva-Jothy (1996) estimated the reproductive success of the damselfly Calopteryx splendens xanthostoma to be 1000 times greater for territorial males. Forsyth and Montgomerie (1987) observed that over the adult lifetime of Calopteryx maculata, territorial males fertilized more eggs per day

(601) than did non-territorial males (15-304). Finally, Tsubaki and Ono (1986)

13

demonstrated that 66% of all copulations for Nannophya pygmaea occurred on the two

top breeding territories (out of thirteen different territory classes).

Odonates, specifically male Leucorrhinia hudsonica, a peatland dragonfly, make ideal

study organisms to explore how spatial distribution of resources influences movement.

Not only are L. hudsonica relatively large, conspicuous, diurnal, and easily handled, but males are territorial and occupy peatland pools (breeding habitats) during the day to look for mates, making them relatively easy to survey.

Examining the processes that may result in particular patterns of animal distribution at fine spatial scales can be methodologically difficult. To examine movement directly, each individual needs to be followed. This is not usually feasible because of technological or logistical issues. Alternatively, one can examine some aspect related to the actual process of movement, such as net displacement. Typically, mark-release-recapture methods are used for such approaches.

My first objective in this chapter is to explore how fine-scale physical characteristics of pools (e.g. pool size, percent cover of emergent vegetation, pH, distance from the pool to the peatland-forest boundary) influence the spatial distribution of male L. hudsonica within peatlands. Because male L. hudsonica require aquatic habitats for part of their life-cycle, at small spatial scales their distribution within peatlands may be structured by processes such as territorial behaviour, selection for optimal breeding sites, and mating behaviour. In addition, because there is an expected gradient of optimal pools within

14

peatlands for mating/ovipositioning/larval development purposes, it is also expected that

there will be a greater number of males on these preferred habitats. My second objective is to explore the distribution of net movement distances of male L. hudsonica. If males are moving randomly, then a random distribution is expected but if movement is constrained (i.e. inhibited) or facilitated by physical or behavioural characteristic(s) within the peatland (e.g. distance between pools, territorial behaviour), then some kind of structured distribution is expected. Because animals respond to the landscape at multiple scales (Senft et al. 1987, Kotilar and Wiens 1990), information gained at the local scale can potentially direct our understanding of ecological dynamics at the broader landscape, where experiments/surveys are much more difficult to conduct (Krawchuk and Taylor

2003).

Methods

Study system

Pools found within peatlands of the Gros Morne Greater Ecosystem (GMGE) are used as oviposition sites for L. hudsonica (see General Introduction) (Holder 2001, McPherson

2003). Males congregate at these pools during warm, sunny, and calm days and may defend territories in order to increase their reproductive success (Hilton 1984).

Mark-release-recapture

A mark-release-recapture (MRR) survey of male L. hudsonica was conducted on a set of three peatlands (B, K, and M) in the Humber River watershed of the GMGE (Universal

Transverse Mercator (UTM) 5500450 0478000; Figure 1.1). These peatlands were

15

chosen based on their relative size (0.44 – 1.07 ha) and easy access. They are a subset of

the peatlands used in the second part of this study (see Chapter 2).

The perimeter of each peatland was mapped using a Garmin E-trex Venture GPS unit

(Datum: NAD 83). UTM co-ordinates were manually recorded, on average, every 4

meters. The area of each peatland was subsequently calculated using the software

ImageTool 3.00 (ImageTool Development Team (2002) (Table 1.1).

In the field, each pool was individually numbered with a surveyor flag. Using the same

GPS as above, I subsequently determined UTM co-ordinates and collected a set of

physical measurements for each pool (Table 1.2). pH was recorded using an Oakton

pHTester 2 meter (with automatic temperature compensation and calibrated daily using

buffer solutions at pH 4.0, 7.0, and 10.0). Estimates of the percent cover of emergent

vegetation (in 5% increments), the slope of the bank (% steep/gradual in 5% increments),

and surface area were also recorded. These variables were related to L. hudsonica larval abundance in previous studies (Holder 2001, McPherson 2003).

On every sunny day between 7 July and 5 August 2003 (the peak flight season for

L. hudsonica) two to six people searched each peatland for male L. hudsonica. Peatlands were surveyed in a systematic fashion (a zigzag path) to ensure an equal amount of search effort in each. Total effort ranged from 1.25 to 19.25 person hours on suitable marking days (22.4 hrs/ha – 62.8 hrs/ha) between 08h30 – 17h00. Individuals were caught with standard aerial insect nets, marked with a unique number on their wing using

16

a Sharpie™ permanent marker, and released at the point of capture. The number of the

pool closest to where they were caught was recorded.

I created two datasets: one with the number of marked males at each pool along with the

associated physical characteristics of the pool, and another consisting of the net distances

that male L. hudsonica moved within the peatland.

Statistical methods

All analyses were conducted using the statistical program R 1.9.0 (R Development Core

Team 2004) and associated packages for spatial analysis (MASS, gstat, and Spatial).

Because biological and ecological processes may lead to an aggregated distribution of

organisms, an issue that arises when conducting field experiments is that observations

may not be independent. Generally, observations from sites (i.e. pools) closer together in

space tend to be more similar than those further apart, resulting in spatial autocorrelation

(Legendre and Legendre 1998, Koenig 1999, Legendre et al. 2002). If such spatial

dependence is not accounted for in statistical models, misleading conclusions can arise

(Koenig 1999).

Semivariograms

I estimated the scale of spatial autocorrelation of occupied pools using semivariograms.

Semivariograms describe the variance between all observations separated by various distances (Kaluzny et al. 1996). They indicate the approximate scale at which pool

17

occupancy was no longer correlated (Kaluzny et al. 1996). All semivariograms were fit

separately for each peatland with a spherical model and using the gstat package in R

(Pebesma 2004).

Ripley’s K

I investigated the degree of aggregation of occupied pools in peatlands using Ripley’s K, which is a measure of how a point pattern process varies with spatial scale (Kaluzny et al.

1996, Crawley 2002). This point pattern analysis tests whether the spatial pattern of occupied pools is independent of the background distribution of all pools. I used this statistic to explore how the distribution of occupied pools (clumped, random, or uniform) varied as I broadened the scale of observation (Cressie 1993) using the Spatial package in

R (Venables and Ripley 2002). Because pool area might be expected to influence occupancy (Holder 2001), I also used Ripley’s K to visually compare the degree of aggregation of occupied pools to the degree of aggregation of pools separated into two size classes (> 5 m2 and ≤ 5 m2) (Appendix 1).

Generalized linear model

To explore how pool-scale characteristics influenced the number of adult male

L. hudsonica at different spatial scales, I used a generalized linear modeling approach

(glm) using poisson and negative binomial families. Negative binomial models were fit

using the function glm.nb (MASS package; Venables and Ripley 2000)) when the

response variable was considered clumped (peatland B and M; mean:variance

ratios >> 1). Fitting a glm with a negative binomial family involves estimating an

18

additional parameter, θ, which is a measure of spatial clumping in the response variable

(Venables and Ripley 2002). θ is bound by zero and positive infinity with values closer to zero indicating an aggregated response (White and Bennetts 1996).

To explore the degree of clustering among males at increasing spatial scales, each peatland in which the distribution of male L. hudsonica was spatially autocorrelated was divided into equal sized grids: 2×2 m, 4×4 m, 8×8 m, 16×16 m, 32×32 m, and 64×64 m

(Figure 1.2). To model these new independent grid units, I calculated or estimated the total number of dragonflies and the surface area of water, the shortest average distance from the pools to the edge of the peatland, the average percent cover of emergent vegetation, the bank slope, and the geometric mean of the pH of the water in each grid at each scale.

For all models, the total number of dragonflies was the response variable and the natural logarithm of the surface AREA1 of the pool, SLOPE of the bank, estimated percent cover of

EMERGENT vegetation, PH, and natural logarithm of the DISTANCE from the pool to the edge of the peatland were fit as potential predictors. At the 32×32 m and 64×64 m scales, for negative binomial models, the true amount of PEATLAND AREA within the grid was

included as a covariate.

A step-wise approach to model selection was used. I first fit models with all pool-level

characteristics at the scale of the pool. I then refit models by keeping variables that

reduced the overall deviance, had good parameter estimates (i.e. small standard errors),

1 Factors included in models are presented in small caps font

19

and appeared to be important in preliminary figures. For data sets that were spatially

autocorrelated, the factors in the final model of the original data set (i.e. at the scale of the

pool) were used to model the peatland at the broader scales (2×2 m to 64×64 m). This

was done so that comparisons among scales could be conducted.

Quantile-quantile plots

I used quantile-quantile plots (q-q plots) with 95% confidence envelopes to compare the

observed distribution of net movement distances for L. hudsonica to the distribution of all

possible movement distances within peatlands B and M (peatland K had too few data

points for meaningful analysis).

Results

General results

A total of 370 adult male L. hudsonica were marked between 7 July and 5 August, 2003

(171 in peatland B, 56 in peatland K, 143 in peatland M). One hundred and sixty re- observations of marked individuals within and between days were made, including multiple resightings of the same individual; 44 in peatland B, 12 in peatland K, and 104 in peatland M (Table 1.1). Gross patterns of occupancy differed among the peatlands: in peatland B 191 out of 443 (43%) of pools were occupied at least once compared to 49 out of 249 (20%) in peatland K and 38 out of 59 (64%) in peatland M. Variance:mean ratios of the total number of dragonflies marked or resighted at each pool were 2.5 in peatland

B, 1.4 in peatland K, and 7.4 in peatland M. The relatively high ratios in peatlands B and

M suggested that the distribution of dragonflies in those peatlands was clumped. Because

20

peatland K exhibited a variance:mean ratio of approximately 1, it was not subjected to

further spatial analysis.

The relatively few individuals caught in peatland K was surprising as the size of the

peatland and the physical data collected for each pool generally fell between the average

measurements for peatlands B and M. However, peatland K did exhibit a slightly higher

pH (4.9 versus 4.3 and 4.2 for peatlands B and M, respectively). Although I did not test

for this specifically, females may prefer not to oviposit in pools with a higher pH because

it may reduce the survival of its eggs/larvae (Corbet 1999). Peatland K also closely

neighboured a much larger peatland to its south (area = 10 724 m2 vs. 34 676 m2) and potentially may have been “feeding” individuals into that area for a variety of reasons

(e.g. more females in the larger peatland). The observed data could also be a result of a variable I didn’t measure.

Aggregation of L. hudsonica at pools

Male L. hudsonica were spatially aggregated around pools at distances up to ca. 60 m in

peatland B (Semivariograms; the ‘range’; Figure 1.3) but not in peatland M. The overall

semivariance in peatland M was considerably higher than in peatland B (Figure 1.3). The

lack of evidence to support spatial autocorrelation among pools in peatland M is likely a

result of the peatland’s small size and overall relatively high occupancy of pools

(Table 1.1).

21

Male L. hudsonica exhibited uniform distributions at fine scales (< 10 m) in both

peatlands B and M (Ripley’s K; Figure 1.4). Distributions of males were aggregated at

broader scales (ca. 70-100 m) only in peatland B. For both peatlands B and M, the

pattern of distribution of occupied pools more closely matched the distribution of large

pools than small pools (Figure 1.5). Sixty-four percent and 95% of large pools versus

19% and 51% of small pools were occupied in peatlands B and M, respectively.

Furthermore, there was an average of 1.7 and 10.9 males on large pools versus 0.25 and

1.0 individuals on small pools in peatlands B and M, respectively. In both peatlands, the

distributions of occupied pools at broader scales tended to be more clumped than was

expected based just on the distribution of larger pools.

The relationship between the average number of male L. hudsonica and pool area for

peatland M had a log-like curve, with approximately 12 to 14 males on pools with a

surface area greater than 15 m2 (Figure 1.6). Because the curve flattened out, it suggested that the peatland may have potentially been saturated with male L. hudsonica. Peatland B displayed an exponential-like relationship (Figure 1.6) suggesting that larger pools had a disproportionately greater number of males than smaller pools and that the peatland was not saturated with L. hudsonica.

22

Numbers of male L. hudsonica related to physical characteristics of the pools

The number of male L. hudsonica at the pool or in the grid significantly increased with

increased AREA of water in all three peatlands at all scales (Table 1.3). Furthermore, the surface AREA of water consistently explained a significant amount of the null deviance in all models (Table 1.4).

It was further observed that grids which had a larger surface area of water tended to have a greater number of larger pools in them (area > 5 m2) and subsequently a greater number of males (Appendix 2). This result supported the original models (the scale of the pool) which suggested that a greater number of male L. hudsonica are found on larger pools.

Estimates of θ increased with increasing grid sizes suggesting that as the scale of observation increases the distribution of male L. hudsonica at pools approaches a random distribution (Figure 1.7). I was unable to estimate θ for two broader-scale models in peatland M, so calculations of θ were conducted using the method of moments estimator

(kMME) (Crawley 2002, Dalthorp 2004). These estimates were consistent with those calculated with glm.nb (Figure 1.7). Lastly, once pool AREA was fit into the models, θ values increased from their respective null models at all scales suggesting that after accounting for the effects of pool AREA distributions of L. hudsonica are more random

(Figure 1.7).

23

Movement and aggregation of L. hudsonica

Net movement distances of male L. hudsonica were more constrained than expected

(given the distribution of pools) beyond scales of ca. 60 m in peatland B and ca. 10 m in

peatland M (q-q plots; Figure 1.8). Within these distances, there is no evidence that males

move non-randomly with respect to pools.

Discussion

Spatial structuring of L. hudsonica

Results from the spatial analysis provide strong evidence that L. hudsonica distributions were spatially structured within peatlands B and M. Two patterns seemed to emerge.

Firstly, males were more uniformly distributed than was expected at finer spatial scales, generally below 10 m. Secondly, movements appeared to be more constrained than expected, at scales broader than ca. 60 m. Although no spatial analyses were conducted on peatland K, as in peatlands B and M, glms suggested that males responded to the structure of large pools.

The observed spatial structure of male L. hudsonica at pools is likely the result of several processes. These include movement, territorial and mating behaviours, and oviposition site selection. The observed distribution of males are consistent with Hilton’s (1984) survey of L. hudsonica that demonstrated that biological processes (i.e. oviposition site selection and territorial behaviour) can influence their spatial distribution. Hilton (1984) observed that male L. hudsonica occupied three particular regions of a small pond (areas

24 with a Sphagnum moss mat) and that mating and oviposition site selection only occurred in these areas.

There are several possible reasons why the relationship between L. hudsonica abundance and the total surface area of water was observed. Adult male L. hudsonica are territorial animals. In a black spruce bog in the Eastern Townships of Québec, Hilton (1984) found individuals defended areas of ca. 1 m2. Although territory size among males may change with population density (Thornhill and Alcock 1983, Kwiatkowski and Sullivan 2002), a larger pool or body of water with a greater surface area would allow more males to set up territories.

Although I was not able to test for this explicitly, I suspect that there were more males on larger pools because females prefer to oviposit in these types of sites (Waage 1987).

Larger pools may potentially have a lower density of predators, allowing an increased survival of eggs or larvae (Blaustein et al. 2004). Furthermore, larger pools tend to have more water in them, and therefore would generally not dry out over the course of a summer (Holder 2001).

Numerous studies on the breeding behaviour of odonates have demonstrated that the sex that chooses the oviposition site varies from species to species (Hilton 1984, Kasuya et al.

1987, Waage 1987, Alcock 1990, Cordero 1995). Michiels and Dhondt (1990) argued that because odonates are r-strategists (Corbet 1980, Ubukata 1981), females will attempt to increase their reproductive success by selecting the best possible micro-habitat for high

25 egg survival and larval development during oviposition. Both Alcock (1987a) and

Waage (1987) conducted manipulation studies that indicated that female Calopteryx maculata are more strongly attracted to sites with the most oviposition resources.

In contrast, Tsubaki and Ono (1986, 1987) found that male Nannophya phygmaea preferred certain breeding sites over others. They later conducted a manipulation experiment and found that males were able to discriminate between attractive and less attractive territorial sites without any mating experience within the study area (Tsubaki and Ono 1995). This suggests that males choose territorial sites by resource quality, rather than by patterns of female dispersion.

The spatial aggregation of male L. hudsonica around large pools may be a result of the extended period of time they spend within peatlands in comparison to females; thus they are able to distinguish between good and poor reproductive sites (Hilton 1984, Tsubaki and Ono 1995). The presence of males at a pool may provide information on pool quality to the females entering the peatland (Switzer 1997b). Conversely, it is possible that female L. hudsonica are able to distinguish between optimal and suboptimal breeding sites, and males, through past reproductive experience or observed reproductive success of their neighbours, are able to predict where females will oviposit and set up territories on those areas (Tsubaki and Ono 1986, Switzer 1997b). The ability of males to obtain information regarding the quality of the breeding site by observing the reproductive success of other males may lead to a positive feedback mechanism where increasingly more males are drawn to these areas. As a consequence, males would aggregate. If males

26

are unable to distinguish good breeding sites, the cost of holding a territory will outweigh

their reproductive gains (Thornhill and Alcock 1983).

Whatever the behavioural reasoning for choosing a particular oviposition site, both glms and Ripley’s K strongly suggest that the observed spatial structuring of L. hudsonica at the scale of the peatland appears to be at least partly influenced by the spatial structuring of large pools within the peatland.

In my models for peatland B, I tested the effect of distance from the pool to the edge of the peatland on dragonfly counts at the scale of the pool, but did not find any significant relationship. This was surprising as preliminary analyses of the data suggested that males were setting up territories on pools found near the edge of the peatland (Appendix 3). I suspected that the distance from the pool to the border of the peatland may have been biologically important in peatland B because of two factors: the reproductive behaviour of females and the poor weather conditions in my study area. Females only briefly appear in the peatland when they are ready to oviposit (Hilton 1984). Furthermore,

Hilton (1984) observed that males would immediate fly towards and attempt to copulate with any female that had entered the breeding site. Compounded with this type of behaviour is the lack of ideal weather conditions for mating in my study system. Because

L. hudsonica are exothermic, they are limited to mating on sunny, calm days, when both sexes venture into the peatland (Hilton 1984). The MRR experiment could only be carried out over eight of the 30 days within their predicted peak flying/mating season

(Holder 2001, McPherson 2003).

27

Because the opportunity to mate is limited by the weather and the relatively infrequent

visits by females to the pools, as well as the short lifespan of adult L. hudsonica between

2 and 4 weeks, the competition among males for females is high. To successfully

reproduce, males must take advantage of suitable mating conditions when they present

themselves. I speculated that males, aware that females come out of the forest and into

the peatlands only when they are ready to mate (Hilton 1984, McPherson 2003), would

distribute themselves over pools greater than 5 m2 (i.e. good oviposition habitats) that are

relatively close to the forest-peatland boundary to get an advantage over other males.

This would potentially allow them to intercept females entering the peatland earlier than other males that had set up a territory further away from the peatland edge.

I did not expect to observe a distance from the pool to the edge of the peatland relationship in either peatland K or M because: 1) there were so few individuals caught in peatland K that males did not need to set up a territory near the peatland-forest boundary due to reduced intraspecific competition, and 2) peatland M was a relatively small peatland; thus all pools were generally found close to the border (median distance between the pool to the edge of the peatland was ca. 4 m and it did not matter how far from the peatland edge males had set up territories. My data are consistent with these expectations.

28

Local scale movement

Although the distribution of distances that L. hudsonica moved in peatlands B and M differed, both peatland populations appeared to move randomly at smaller spatial scales

(60 m and 10 m for peatlands B and M, respectively) while at larger distances their movement appeared to be more constrained than expected. Movement may have been constrained at these scales because males may have had an affinity to sites where they were last successful in mating (Thornhill and Alcock 1983, Switzer 1997a). Therefore, moving to an unknown area would not be advantageous to the reproductive success of the individual.

It has been shown that some birds (Nager et al. 1996, Kokko et al. 2004) as well as at least one odonate species (Switzer 1997a) display a “win-stay, lose-switch” strategy

(Switzer 1993). That is, individuals that experienced a mating failure leave the site in which they were unsuccessful despite the potential costs of moving to a new site (e.g. energetic costs, increased mortality risks, no knowledge of the resource quality of the new site) (Larsen and Boutin 1994). Switzer (1997a) also demonstrated experimentally that males, which voluntarily moved sites, moved to higher-quality territories.

Odonates are strong flyers, with many able to fly tens of kilometers and some with the ability to make long migrational journeys up to thousands of kilometers (Corbet 1999).

Although no dispersal surveys have been formally conducted on L. hudsonica, males have the ability to travel at least 1.6 km (see Chapter 2). Therefore, these relatively small

(constrained) net movement distances observed, as well as the uniform distribution of

29 occupied pools at small scales (ca. 0-10 m), suggested that there was a process (e.g. territorial behaviour) that was influencing their movement. If males were holding territories, I would have expected a uniform distribution of their net movement distances.

Although I did not collect information on their daily activity budget or exclusively record defensive behaviour for individuals (classic measurements of territorial behaviour; Gill

2000), the random distribution of net movement distances at small spatial scales suggested territorial behaviour in these animals. Males moving within these distances may have been defending multiple territories (i.e. a series of satellite pools around a larger pool) (Conrad et al. 1999). If males were able to hold more or larger territories, they may have been more successful in mating. Lastly, after copulation, females may choose to oviposit in a series of pools and males may guard them as they oviposit.

Ovipositing in several pools may contain several advantages for the survival of the egg and larvae. These include a lower larval density which may result in lower larval competition and a decrease in the risk of disease and predator attraction (Tsubaki and

Ono 1995).

30

N

Water

Figure 1.1. Aerial photograph of the study area. The landscape contains peatlands, forest, scrub, small bodies of water, and commercially harvested areas (“cut”). Peatlands B, K, and M are identified with letters over the area. Photograph courtesy of Corner Brook Pulp and Paper Ltd.

31 01050 55

Unoccupied pool Occupied pool 000 5501 ng i h t r no M T U 950 5500 0 550090 478000 478050 478100 478150 478200

UTM easting

Figure 1.2. A schematic of peatland B divided into 32×32 m grids - an example of how the peatlands were randomly divided into grids. Open circles represent pools in which no L. hudsonica were found. Filled circles represent pools in which L. hudsonica were caught.

32

Peatland B Peatland M

50

40

1.0 e

c 30 n a i r a v i m Se

20

0.5

10

0.0 0 0204060 01020304050

Distance (m)

Figure 1.3. Semivariogram of all occupied pools in peatlands B and M using a spherical model. Distance between pairs of pools is plotted against semivariance (the variance between data points).

33 0

. clumped 1

Peatland B Peatland M 8 . 0 .6 0 4 . 0 .2 0 Proportional deviation from the expected value 0 .

0 uniform

0 20406080100

Distance (m)

Figure 1.4. Deviations from the expected distribution for occupied pools in peatlands B and M. Solid lines show the upper and lower 95% confidence intervals (CI) from 1000 Ripley’s K simulations. Points that fall above the upper 95% CI are clumped, below the lower 95% CI are uniform, and in between the 95% CI are randomly distributed. Distance is the paired distance between pools.

34 y distributed. l ) in peatlands B 2 lations. Points that u clumped uniform

0 CI are random 8 2 2 d 5m 5m e v r > ≤ a a se e e 0 Ob Ar Ar 6 1000 Ripley’s K sim nd M ) and large pools (area > 5 m 0 a 2 4 l

5m Peat ≤ 0 ea , and in between the 95% r ls (a form 02

ce intervals (CI) from

.0 .0 .0 1 1 1 8 8 8 . . . 0 0 0 .6 .6 .6 0 0 0 4 4 4 . . . 0 0 0 .2 .2 .2 0 0 0 0 0 0 . . . 0 0 0 r small poo o Distance (m) tion f 100 u strib i 0 8 d 0 6 nd B a l at 2 2 0 d 4 the upper and lower 95% confiden the expected 5m 5m e Pe v > ≤ r e a a s e e 0 Ob Ar Ar 2

0

.0 .0 .0 1 1 1 8 8 8 . . . 0 0 0 .6 .6 .6 0 0 0 4 4 4 . . . 0 0 0 .2 .2 .2 0 0 0 0 0 0 . . . 0 0 0

. Deviations from

om the expected value value expected the om fr deviation Proportional and M. Solid lines represent fall above the upper 95% CI are clumped, below uni Distance is the paired distance between pools.

Figure 1.5 35

Peatland B Peatland M

8

6 10 s l a u d i v di in of

r 4 e b m nu an e 5 M

2

0

0 0 5 10 15 20 25 0 5 10 15 20 25 Pool area (m2)

Figure 1.6. Locally-weighted regressions lines showing the influence of pool area on the mean number of individuals found on pools in peatlands B and M.

36

A B C 2222 1111 3.03.0 Peatland B - null model

0000 Peatland B - model of best fit 1111 Peatland M - null model Peatland B Peatland M - model of best fit 0000 .5.5 1111 22 Peatland M Peatland B - glm.nb Peatland B - kMME Peatland M - glm.nb

8888 Peatland M - kMME 8888 2.02.0 a

.5.5

6666 11 het T 46464646 4444 1.01.0 .5.5 00 02020202 02020202 0.00.0

pool 2x2 4x4 8x8 16x16 32x32 64x64 pool 2x2 4x4 8x8 16x16 32x32 64x64 pool 2x2 4x4 8x8 16x16 32x32 64x64

Scale (m)

Figure 1.7. Comparison of θ values across 7 spatial scales for peatlands B and M for A) the null models, B) the glm.nb and methods of moment estimator (kMME), and C) the null models and models of best fit. Missing θ values in A, B, and C are due to a θ that was not estimable.

37

Peatland B Peatland M 120 120 100 100 ) 0 0 m ( 8 8 ved er 0 0 obs 6 6 es anc t s i 0 D 0 4 4 0 2 0 2 0 0

0 50 100 150 200 0 50 100 150

All paired distances (m)

Figure 1.8. Quantile-quantile plots of observed net L. hudsonica movement distances (open circles) in peatlands B and M with 95% confidence limits (solid lines). Circles that fall outside of the 95% confidence envelopes indicate that movement is not random.

38

Table 1.1. Physical characteristics and mark-release-recapture results for peatlands B, K, and M between 7 July and 5 August, 2003.

No. of recaps Peatland No. of No. of Peatland No. of No. (w/in and Effort Peatland perimeter occupied sample area (m2) pools caught btw days) (mins/m2) (m) pools (%) days (%)

B 10729 717 443 191 (40%) 171 44 (25%) 5 0.309

K 10724 638 249 49 (20%) 56 12 (20%) 2 0.134

M 4366 536 59 38 (60%) 143 104 (67%) 6 0.377

Table 1.2. Summary of the mean physical measurements (or geometric mean) taken from pools in peatlands B, K, and M (standard errors in parentheses). Included is the average shortest distance between the pools and the edge of the peatland.

Emergent Slope Peatland Area (m2) Distance (m) pH vegetation (%) (% steep) B 3.5 (±0.38) 8.5 (±0.28) 17.1 (±1.1) 92.1 (±1.0) 4.28 (±0.0097)

K 7.2 (±1.3) 12.9 (±0.52) 11.4 (±0.86) 92.5 (±1.2) 4.94 (±0.013)

M 20.7 (±11.2) 4.1 (±0.36) 6.6 (±0.90) 81.9 (±3.9) 4.24 (±0.018)

39

Table 1.3. Parameter estimates and standard errors for co-variates (the surface AREA of water of the pool or grid and the amount of LAND.AREA in each grid) of the glm.nb models across 7 spatial scales (peatlands B and M) and the glm model of best fit (peatland K). Models would not converge at the 64×64 m resolution for peatland M.

Peatland Scale Co-variate Estimate SE P(>|z|) pool AREA 0.644 0.0754 < 0.001 2×2 AREA 0.632 0.0745 < 0.001 4×4 AREA 0.55 0.0798 < 0.001 8×8 AREA 0.487 0.0916 < 0.001 B 16×16 AREA 0.668 0.139 < 0.001 AREA 0.893 0.209 < 0.001 32×32 LAND.AREA 0.000702 0.000436 0.107 AREA 0.992 0.383 0.00953 64×64 LAND.AREA 4.45E-05 0.000262 0.865 pool AREA 1.97 0.697 < 0.001 2×2 AREA 1.97 0.697 < 0.001 4×4 AREA 2.04 0.732 < 0.001 M 8×8 AREA 2.23 0.788 < 0.001 16×16 AREA 3.03 1.3 < 0.001 AREA 2.34 1.34 0.0642 32×32 LAND.AREA 0.00115 0.00201 0.567 K - AREA 0.016 0.002 < 0.001

40

Table 1.4. Residual deviances of the glm.nb models across 7 spatial scales (peatlands B and M) and the glm model of best fit (peatland K). Covariates include the surface AREA of water of the pool or grid and the amount of LAND AREA in each grid. Models would not converge at the 64×64 m resolution for peatland M.

Residual Residual Peatland Scale Co-variate p(>|Chi|) deviance df NULL 411 442 pool AREA 317 441 < 0.001 NULL 400 410 2×2 AREA 308 409 < 0.001 NULL 324 312 4×4 AREA 263 311 < 0.001 NULL 200 159 8×8 B AREA 163 158 < 0.001 NULL 94.5 57 16×16 AREA 66.7 56 < 0.001 NULL 63.2 20 32×32 AREA 26 19 < 0.001 LAND.AREA 23.3 18 0.098 NULL 61.5 7 64×64 AREA 9.42 6 < 0.001 LAND.AREA 9.39 5 0.867 NULL 152 57 pool AREA 64.7 56 < 0.001 NULL 152 57 2×2 AREA 64.7 56 < 0.001 NULL 146 55 4×4 AREA 63.1 54 < 0.001 M NULL 121 46 8×8 AREA 51.5 45 < 0.001 NULL 75.3 23 16×16 AREA 27.2 22 < 0.001 NULL 19.9 8 32×32 AREA 11.3 7 0.003 LAND.AREA 10.9 6 0.534 NULL 227 248 K AREA 202 247 < 0.001

41

Chapter 2. Effects of three matrix types on between-peatland connectivity for the dragonfly Leucorrhinia hudsonica in western Newfoundland.

Abstract

I investigated effects of clearcutting on landscape-scale movement behavior of a peatland dragonfly (L. hudsonica) in western Newfoundland, Canada. I conducted a mark-release- recapture experiment to determine if movement rates between peatlands are influenced by intervening matrix (forest, clearcut, or scrub), or other landscape (distance) or peatland features (surface area of water, pH, percent cover of emergent vegetation, slope of the pool banks). According to multi-strata models that were fit in Program MARK using a quasi likelihood approach, the average percent of males moving among peatlands was 1.2% (n = 1280 marked individuals), with significantly more emigration out of peatlands having a smaller surface area of water. Matrix types were included in two of the four best models explaining inter-peatland movement but the 95% confidence limits of estimated coefficients overlapped with zero. Parameter estimates consistently suggested that forest matrix impedes movement more than cut or scrub matrices and that scrub matrix facilitates movement. Generalized linear models with a binomial distribution suggested that, at short distances, a greater proportion of males moved between peatlands separated by forest than non-forest but that at larger distances a greater proportion moved through non-forest. Competition for territories (i.e. pools) in peatlands with less water may be greater, thus compelling some males to emigrate. Furthermore, movement out of peatlands across different types of matrix habitat may not be a simple

function of the permeability of that habitat, but instead, may be a function of the

interaction of different behaviours being invoked at different spatial scales.

42

Introduction

Most animals have the ability to move through space to some degree. Movement

behaviours are generally linked to immediate threats or needs. For example, an

individual may move in order to forage (Ward and Saltz 1994), access habitat (Dunning

et al. 1992), escape unfavourable conditions (McIvor and Odum 1988, Holt 1990), or find

mates (Hilton 1984). Empirical studies have shown that movements of individuals are

typically highly localized throughout their lifetime (Harrison 1989, Roland et al. 2000,

Purse et al. 2003, Botero-Garcés and Issacs 2004). However, rare long distance dispersal

movements can have important ecological implications for populations. For example,

these long distance movements can promote colonization of new habitats and gene flow,

thereby increasing the long-term survival of a species (Gustafson and Gardner 1996,

Simpkin et al. 2000). Understanding movement patterns at relatively large spatial scales

provides essential insight on landscape connectivity (Ricketts 2001), patch boundary

dynamics (Schtickzelle and Baguette 2003), spread of pest populations (Botero-Garcés

and Issacs 2004), and metapopulation dynamics (Hanski and Thomas 1994).

An individual moving to access resources in the landscape is not only strongly influenced

by its physiology and behaviour (Ims 1995, Mauremooto et al. 1995, McIntyre and Wiens

1999) but also by the intervening habitat that separates resource patches (i.e. matrix)

(Goodwin and Fahrig 2002, Jules and Shahani 2003, Matter et al. 2003). Landscapes are heterogeneous at multiple spatial and temporal scales, and movements reflect how organisms responds to this heterogeneity (Kotilar and Wiens 1990, Wiens et al. 1997,

Roland et al. 2000). The degree to which a moving organism is impeded or facilitated by

43 the matrix is termed “connectivity” (Taylor et al. 1993). Connectivity is a critical element of landscape structure, having implications for ecological processes such as population dynamics, gene flow, and predator-prey interactions (Mauremooto et al. 1995,

Mech and Hallet 2001). An animal’s decision to move through the matrix may be based on perceived costs of movement (e.g. mortality risks), its present needs, or the current condition within the resource patch it occupies (e.g. resource levels, population density, physiological stress levels) (Larsen and Boutin 1994, Bowler and Benton 2005). When resources are limited, predation risk is high, or intraspecific competition is greater than the potential cost of crossing the matrix, conditions favour movement (Hein et al. 2003,

Russell et al. 2003).

Forest harvesting and other human activities are rapidly altering landscapes through changes to the matrix. Loss of natural habitat has been shown empirically to influence between-patch movement rates for many organisms beyond that of simple Euclidean distance alone (Gustafson and Gardner 1996, Roland et al. 2000, Chardon et al. 2003).

The matrix therefore has the potential to affect population dynamics by altering the abundance and distribution of populations at multiple spatial and temporal scales (Crist et al. 1992, Jonsen et al. 2001, Betzholtz 2002, Chardon et al. 2003). To understand how anthropogenic disturbances may influence animal processes and population dynamics, it is often important to quantify matrix effects on movement rates (Roland et al. 2000,

Ricketts 2001).

44

Matrix containing low -value or degraded habitat for a particular species (e.g. clearcuts

and agricultural landscapes for forest species) can increase mortality risk (St. Clair et al.

1998, Hanski et al. 2000, Purse et al. 2003), impede movement (Fahrig and Merriam

1985, Pither and Taylor 1998, Jonsen et al. 2001, Hein et al. 2003), or facilitate

movement (Pither and Taylor 1998, Roland et al. 2000, Ricketts 2001).

It is imperative that in anthropogenically altered landscapes connectivity within landscapes is maintained for the variety of species that inhabit an area. Excessive rates of movement may be detrimental to the survival of a population by promoting spatial synchrony (sensu stricto; (Sutcliffe et al. 1997, Earn et al. 2000) in population dynamics and increasing the likelihood of regional population collapse by eliminating the opportunity for demographic rescue (Brown and Kodric-Brown 1977, Liebhold et al.

2004). Conversely, isolation of subpopulations through reductions in movement can also prevent demographic rescue and promote inbreeding or restrict gene flow (Lande et al.

1999, Frankham et al. 2003). In both of these cases, some populations may become more vulnerable to extinction (Earn et al. 2000, Frankham et al. 2003). Improving our understanding of long-distance movements in altered landscapes is thus important to gaining a better understanding of how anthropogenic changes might influence demography.

Studies examining the connectivity of the landscape must be performed at an appropriate spatial scale - the scale at which the species of interest is capable of moving (Jackson

1991, May 1994). However, as the mobility of a species increases, it becomes

45 increasingly more difficult to conduct empirical movement studies for technological and/or logistical reasons (Koenig et al. 1996, Krawchuk and Taylor 2003). Studying insects is advantageous because their movement abilities tend to be smaller than most other animals (e.g. fish, birds, mammals) making it easier to conduct field studies at the

(perceived) appropriate spatial scales. Although we cannot use the information gained to directly predict how other species will be affected by the landscape, insect studies give us a preliminary understanding of the important factors to consider when conducting future studies on other species (Wiens et al. 1993).

Much is known about basic life history processes of odonates (e.g. territorial, mating, and defensive behaviours) (Hilton 1984, Waltz and Wolf 1988, Tsubaki and Ono 1995,

Thompson 1997, Switzer 2002). However, despite a growing interest in odonates world- wide and a general decline in their populations throughout North and South America,

Europe, Asia, Australia, and Africa (Corbet 1999, Purse et al. 2003, IUCN 2004, Watts et al. 2004), only recently have studies been undertaken to explore odonate movement in a landscape context (Switzer 1997b, Pither and Taylor 1998, Conrad et al. 1999, Jonsen and Taylor 2000b, Jonsen and Taylor 2000a, Holder 2001, Angelibert and Giani 2003,

McPherson 2003, Purse et al. 2003).

The purpose of this chapter is to develop an understanding of how the landscape in the

Humber River watershed of western Newfoundland, which has a matrix comprised of forest, scrub, and clearcut vegetation covers, influences inter-patch connectivity for

L. hudsonica, a peatland dragonfly, at relatively large spatial scales. Specifically, I

46 wanted to quantify the temporal frequency and spatial extent of movement and determine whether matrix type influences movements of L. hudsonica between habitat patches

(peatlands). To explore these topics, I undertook a large-scale mark-release-recapture survey. Because it has been shown experimentally and empirically that altered landscape structure influences movement abilities for many organisms (Roland et al. 2000, Jonsen et al. 2001, Ricketts 2001) including odonates (Pither and Taylor 1998, Jonsen and

Taylor 2000b, Purse et al. 2003), I expected to observe differences in movement rates through the three matrix types. This prediction is supported by past surveys in this study system, which demonstrated an effect of matrix on incidence and abundance of L. hudsonica larvae (Holder 2001, McPherson 2003).

Methods

Study system and data collection

The Gros Morne Greater Ecosystem (GMGE) is a system comprised of a mosaic of forest, peatland, scrub (stunted black spruce, ca. 1.5-2 m in height), bodies of water, and clear-cut patches. L. hudsonica require both peatland and forest habitats during their adult life cycle. Peatlands are used as mating grounds and for ovipositing and larval development, while forest habitats are used for roosting, foraging, and for sexual maturation of tenerals (Hilton 1984, Corbet 1999).

A mark-release-recapture (MRR) survey was conducted on a set of twelve peatlands

(which were divided into 3 sections – 4 “northern”, 4 “central”, and 4 “southern” peatlands) in the Humber River watershed between 7 July and 5 August 2003, the peak

47

flight period of L. hudsonica (Holder 2001, McPherson 2003) (Figure 2.1). All peatlands

were situated within an area of approximately 2 km2 and the maximum distance between two peatlands was approximately 1.7 km (Table 2.1). Peatlands were chosen to represent a range in size, accessibility, configuration within the landscape, and amount and type of intervening matrix found between pairs of patches.

On every sunny day between during the sampling period, between one and eight people searched each peatland for male L. hudsonica. Dragonflies were caught with standard insect aerial nets. Each individual was marked with a unique number on its wing using a

Sharpie™ permanent marker, and released at the point of capture. Peatlands were surveyed in a systematic fashion (a zigzag path) to standardize search effort based on the area of each peatland. Total effort ranged from 1.25 - 21.5 person hours on suitable marking days (18.3 h/ha – 67.7 h/ha), with sampling being conducted between 08h30 –

17h00.

The perimeter and area of each peatland was measured with a Garmin E-trex Venture

GPS unit (Datum: NAD 83). Physical measurements were obtained for each pool in peatlands B, K, and M, and on average for 30% of all pools in the remaining nine peatlands. I measured or estimated a set of physical variables for each sampled pool. pH was recorded using an Oakton pHTester 2 meter (with automatic temperature compensation and calibrated daily using buffer solutions at pH 4.0, 7.0, and 10.0).

Estimates of the percent cover of emergent vegetation (5% increments), slope of the bank

(% steep/gradual in 5% increments), and surface area of water were also recorded. These

48

variables were statistically associated with to L. hudsonica larval abundance in previous

studies (Holder 2001, McPherson 2003).

Distance and amount of intervening matrix type between peatland pairs were measured

using ImageTool 3.00 software (ImageTool Development Team 2002). Using a

conservative approach, the closest distance between each possible pair of peatlands was

measured from a georeferenced aerial photograph of the study system. Area of the

matrix was determined by drawing lines from the widest points between pairs of

peatlands and measuring the area of each matrix type in between them. The proportion of

each intervening matrix type was calculated and then multiplied by the distance between

the two peatlands to get an estimate of the distance of each matrix type that the animal

must traverse when moving between a pair of peatlands.

Statistical analysis

Even though long distance movements are difficult to document (Purse et al. 2003), it

was primarily these broad scale movements of male L. hudsonica that were the focus of this study. Consequently, for the analyses, I grouped peatlands into six sections (peatland

A, B, H, I, M, and O) (Figure 2.1). The grouping of peatlands was decided by measuring the type and amount (m) of matrix that the individual would have to move through in order to reach a peatland that was not in its section. If the amount of matrix type between two peatlands in the same section and another peatland in a different section (e.g. peatland A to I and peatland B to I, or peatland D to I) was greater than 200 m and within ca. 20% from each other, then the peatlands were grouped. Grouping of peatlands within

49

sections meant that all dragonflies within each group were amalgamated into one

“peatland” for analytical purposes.

To measure how matrix type affected movement rates between peatlands, I fit multi-

strata models for live recaptures using a likelihood approach in program MARK 5.1

(White and Burnham 1999). This approach provides estimates of daily movement rates

(ψ) between peatlands while taking into account daily survival (Φ) and recapture rates (p)

for each stratum (i.e. peatland or group of peatlands). A suite of potential predictor

variables that could influence movement rates between functional groups of peatlands

included the natural logarithms of the amount of CUT1, FOREST, and SCRUB MATRIX, the natural logarithm of the shortest DISTANCE between peatlands, the average PH of pools, the

average percent cover of EMERGENT vegetation of each pool, and the natural logarithm of

the total surface area of WATER. Because patch size can be functionally defined as the amount of resource available (Matter et al. 2003), the total surface area of water within a peatland (i.e. breeding habitat) was used as a measure of patch size rather than peatland size per se.

A candidate set of ten models was developed a priori based on prior knowledge of the system (Holder 2001, McPherson 2003). Included in these were models specifying physical characteristics of the peatland pools and terms for the amount of FOREST, CUT, and SCRUB MATRIX between pairs of peatlands, as well as models that included the same physical pool characteristics plus DISTANCE between peatlands. Comparison of these

1 Factors included in models are presented in small caps font

50

models allowed an evaluation of the importance of the matrix. A median c-hat goodness

of fit test (GOF) was conducted on the global model to assess the amount of variance in

the data. MARK returned a c-hat value which was then used to correct Akaike’s

Information Criterion (AIC) values to quasi-likelihood adjusted AIC (QAICc), which is

calculated as the likelihood divided by the amount of overdispersion in the data (Cooch

and White 2005). Models were then ranked using QAICc.

Finally I also used a generalized linear model (glm) with a binomial distribution to

explore the relationship between matrix type, distance, and movement. I modeled the

number of animals that moved versus the number that didn’t move as a function of the

interaction between both the DISTANCE and the amount of intervening FOREST MATRIX between peatlands.

Results

A total of 1280 individual male L. hudsonica was marked between 7 July and 5 August,

2003. Including multiple re-sightings of the same individual, 220 (17.2%) re- observations of marked individuals were made and of those recaptured, 25 (11.4%) had moved between functional groups of peatlands (Table 2.1). The longest movement observed was ca. 1560 m. Of those individuals making inter-peatland movements, the average movement distance was 432 ± 78 m (mean ± SE). There was a general decline in the number of male L. hudsonica that moved between peatlands as intervening distance

increased (Figure 2.2).

51

Effect of matrix and distance on movement rates

GOF tests suggested a value of 1.8665 for c (indicating overdispersion). Daily apparent

survival rate for male L. hudsonica was 0.87 (±0.013), daily recapture rate was 0.15

(±0.033), and the average daily movement rate (the probability that an individual animal

would leave any peatland) was 0.012 (±0.0033) (Table 2.2).

Two of the four top MARK models included MATRIX type (forest, cut, and scrub matrices;

Table 2.3). The 95% confidence interval of all three parameter estimates for MATRIX types included zero (Table 2.4). Similarly, 95% confidence limits for parameter estimates of the two top models containing a DISTANCE term also included zero. The poor estimates may be a function of sparse data (i.e. having found very few individuals to have moved between peatlands) but could also be a result of a more complex (i.e. non-linear) relationship between matrix and distance (see Interaction between matrix type, below).

Parameter estimates for matrix variables were, however, consistent across all models, with FOREST MATRIX impeding movement more than an equivalent distance through CUT

MATRIX, and SCRUB MATRIX facilitating movement.

Influence of non-matrix variables on movement rates

The total surface area of water in the peatland strongly influenced male L. hudsonica movement; the amount of water was inversely related to dragonfly emigration in all models. Although the average percent cover of EMERGENT vegetation in pools was

present in two of the top four models, 95% confidence limits for parameter estimates

included 0.

52

Interaction between matrix type and distance between peatlands

In the glm, where I evaluated the relationship between DISTANCE and FOREST MATRIX on movement, parameter estimates for DISTANCE, FOREST MATRIX, and the interaction between

DISTANCE and FOREST MATRIX were significant (p < 0.05) (Table 2.5). Both DISTANCE and

the interaction between DISTANCE and FOREST MATRIX also contributed significantly to the fit of the model (Table 2.6).

The interaction plot between DISTANCE and amount of FOREST MATRIX indicated that, at

short distances, there was a higher proportion of movement between peatlands separated

by forest than those separated by non-forest matrix (CUT and SCRUB). However, at larger

distances, the opposite trend was observed: there was a higher proportion of movement in

non-forest matrix (Figure 2.3). This interaction suggests that the relationship between

movement and matrix is more complex than I had initially thought.

Discussion

Influence of the matrix on inter-peatland movement

The results suggest that both landscape structure and total surface area of water in a

peatland affect inter-peatland movement of male L. hudsonica. Furthermore, they also

suggest that the type of intervening matrix influences movement beyond that expected

from distance between patches alone, and potentially in a non-linear fashion.

The influence of matrix composition on connectivity between peatlands has important

implications for studies that assume that patch isolation is independent of landscape

53 structure (Fahrig and Paloheimo 1986, Harrison 1989, Conrad et al. 1999, Baguette et al.

2000). In the present system, failing to explore the effect of matrix type on movement rates would not have revealed that distance and matrix type influence movement of L. hudsonica in a non-linear fashion. Although metapopulation theory asserts that patch isolation is a function of distance between two patches (Hanski and Simberloff 1997), in this study, matrix was also an important factor affecting patch isolation for male

L. hudsonica.

Albeit with poor parameter estimates, the multi-strata models suggested that forest matrix was more resistant to movement than cut and scrub matrices and that scrub matrix increased connectivity between peatlands. These findings were supported by prior work on L. hudsonica in the GMGE (McPherson 2003). The lack of movement data could be the result of the difficulty in observing longer-distance movement events (Harrison 1989,

Koenig et al. 1996, Conrad et al. 1999). As distance between patches increases, the area in which an individual could potentially move increases exponentially (Purse et al. 2003) making such studies inherently difficult.

Previous broad-scale studies on insect movement have shown an effect of matrix type on landscape connectivity. Ricketts (2001) demonstrated that for four species of butterflies, conifer matrix was 3-12 times more resistant to movement than willow matrix. Goodwin and Fahrig (2002) tracked individual goldenrod beetles (Trirhabda borealis) in three different matrix types: goldenrod, cut vegetation, and cut vegetation containing camouflage netting and demonstrated that beetle movement behaviour was a function of

54

matrix type. Also, Jonsen and Taylor (2000b) conducted a manipulative experiment in

which they compared the movement behaviour of a damselfly species away from streams

(i.e. breeding sites) in forested, partially forested, or non-forested landscapes. They

found that the tendency to move away from streams and rates of net displacement

differed among landscape types.

Patterns of movement by L. hudsonica at different scales is likely the result of different behaviours. At the fine scale, at which foraging operates, there is an increased likelihood of L. hudsonica traversing the peatland edge when the peatland borders forest because they forage in forest ((Jonsen and Taylor 2000b). However, a comparable behavioural response to a cut edge bordering a peatland is less likely because resource availability is limited in the cut. Longer distance movements are likely a consequence of behaviours other than foraging. L. hudsonica may emigrate from a peatland to increase mating opportunities, or to avoid competition for mates or territories (Chapter 1). As a consequence, I speculate that movement in each matrix type results from different behaviours, and that these behaviours interact with the structure of the landscape to generate large-scale patterns of movement.

Because cut matrix had little structural complexity and generally could not be exploited for food, shelter, or mating opportunities (Rith-Najarian 1996), I suggest that it would elicit a behavioural response from males which would result in rapid, straight-line movement (Root and Kareiva 1984). Forest matrix, however, is structurally complex and offers important resources for L. hudsonica (e.g. roosting and foraging opportunities)

55

(Hilton 1984, Corbet 1999). Thus, it may inhibit between-patch movement rates by compelling individuals to remain in the desirable area (Yang 2000). I speculate that the higher proportion of animals that moved through forest matrix at short distances (0-100 m) might be foraging males who are prone to making many fine-scale decisions (e.g. tight turn angles (Corbet 1999)) and consequently just ‘end up’ in a new peatland. Such speculation is consistent with behaviors suggested by Jonsen and Taylor (2000a), and is also consistent with my observations of constrained within-bog movement in Chapter 1.

It is also possible that male L. hudsonica, which use sunlight as visual cues, are disorientated in the forest because of the abundance of shade provided by trees (Corbet

1999). An inadequate amount of sunlight would promote unintended short movements and long distance movements requiring strong orientation would be inhibited.

Structurally, scrub matrix was an intermediate between cut and forest matrix. Though trees were generally stunted, this matrix type still provides some resources to L. hudsonica (e.g. shelter, protection from predators). Because scrub matrix is dense, males likely travel above the trees while moving between peatlands. In such situations, if weather turns unfavourable or they sense a predator, individuals can drop into the trees for protection. In this way, movement would be facilitated through this matrix type because individuals would perceive a lower risk compared to similar movements through cut matrix.

Lastly, an increased perceptual range in both cut and scrub matrices may have contributed to the higher connectivity between peatlands than in forest matrix. The lack

56 of structural interference in the cut matrix and the flight behaviour in scrub matrix (i.e. above the trees), may have given individuals better information about their surroundings and actively allowed them to make decisions in the direction they wanted to move

(Zollner 2000). Because long-distance movement is usually deliberate, as opposed to accidental, it would be ideal to move through a matrix type that does not obstruct long- distance vision.

Extrapolating the influence of matrix type by male L. hudsonica counts

Joly et al. (2001) contended that the effect of the matrix could be estimated, indirectly, by comparing the abundance of individuals in habitat patches that were surrounded by differing amounts of matrix types. McPherson (2003) found that the incidence and relative abundance of L. hudsonica larvae in peatlands surrounded by scrub matrix was significantly higher than in those surrounded by forest matrix. In addition, she noted that the proportion of occupied pools (i.e. pools in which L. hudsonica larvae were found) was higher in cut landscapes than forested landscapes (McPherson 2003). Similarly, in my study area, an increased number of male L. hudsonica was observed in peatlands surrounded by cut matrix than by forest matrix (178 males/ha and 115 males/ha, respectively). The results suggest potentially higher rates of movement through cut than forest matrix. Haddad and Baum (1999) demonstrated in an experiment with three habitat-restricted butterfly species that population densities were higher in habitat patches that were connected by corridors than in similar patches that were not connected. They further suggested that these greater densities were likely a result of a higher movement rate.

57

Influence of non-matrix variables on interpeatland movement

Because L. hudsonica defended territories and bred at pools, competition among males

may have been higher in peatlands that had a lower surface area of water. Since finding a

mate is critical for all sexually reproducing organisms, if an individual was unable to find

a mate or hold a territory, or if the costs of holding a territory outweighed the

reproductive benefits, the individual would have been pressured into emigrating despite

the energy costs and potential risks of dispersing across the matrix and finding a new

habitat patch (Switzer 1997a, Schtickzelle and Baguette 2003). Previous studies have

had shown similar findings, where individuals emigrated from smaller habitat patches

because of the lack of available resources, to larger ones in hopes of improved

reproductive success (Cronin 2003, Schtickzelle and Baguette 2003).

Although emergent vegetation did not significantly influence L. hudsonica emigration,

previous work on odonates demonstrated that females preferred to oviposit in pools with relatively greater amounts of vegetation (Waage 1987, Corbet 1999). This may have been related to habitat needs of larvae. One of the greatest threats to young larvae is predation (Corbet 1999), and females that oviposited in pools having a relatively large amount of emergent vegetation could have favoured increased survival of their offspring by selecting sites that afforded larvae with escape cover (Waage 1978, Waage 1987,

Corbet 1999, Needham et al. 2000).

58

Lastly, the relationship between distance and the proportion of animals moved suggests

that emigration generally decreased with increasing distance. This followed the expected

relationship from scores of studies (Koenig et al. 1996, Conrad et al. 1999, Baguette et al.

2000). However, despite the relatively large distances separating the north and south

peatlands, seven of the 25 individuals (28%) moved from one end of the study system to

the next. This constitutes strong evidence that these animals have the ability to move

even greater distances and that even this large-scale study was unable to capture the full extent of movement within the species.

The relatively high rate of movement between patches and high proportion of animals moving between the south to the north peatlands suggests that male L. hudsonica in this system exist as a “patchy population” as opposed to a “metapopulation” of isolated subpopulations (Harrison 1991). It appears that L. hudsonica are well-adapted to the patchy distribution of peatlands in the GMGE because of these high rates of movement.

Although the connectivity of the Humber River watershed appears to be influenced by matrix type, peatlands are functionally linked at the scale in which I conducted this survey. Clear cuts were not barriers to movement and L. hudsonica may be responding to changes in the physical environment by shifting their distribution. However, longer-term and larger-scale data sets (replicated in different landscapes) would be needed to evaluate changes in population demographics so that the influence of matrix types on the process of movement and consequently on population dynamics could be determined.

59

N

Water

Figure 2.1. Aerial photograph of the study landscape. The area is composed of forest, scrub, peatlands, small bodies of water, logging roads, and commercially harvested areas (“cut”). Peatlands A, B, D, E, H-O are identified with letters over the area. Grouping of similar peatlands are circled and the name of the functional peatland is identified with a star ( ). Photograph courtesy of Corner Brook Pulp and Paper Ltd.

60

0.03 d e v o m 0.02 les a m on of i t or op r P 0.01

0.00

0 500 1000 1500 Distance (m)

Figure 2.2. Locally weighted regression lines showing the influence of distance between peatlands on the proportion of animals moving.

61 030 0.

0-50% forest

025 51-100% forest 0. d e v o 020 m s 0. le a m 015 on of 0. i t opor r P 010 0. 005 0. 000 0.

0-700 701-1600

Distance class (m)

Figure 2.3. Interaction plot of the proportion of male L. hudsonica moving between peatlands in two distance classes. Closed triangles (with standard error bars) represent peatlands that are separated by more than 50% FOREST MATRIX. Open circles (with standard error bars) represent peatlands that are separated by less than 50% FOREST MATRIX (i.e. separated by mainly CUT and SCRUB MATRIX).

Table 2.1. Summary of the average physical measurements and results from the MRR survey on the 12 sampled peatlands in the Humber River watershed in 2003 (standard errors in parentheses).

No. Peatland Slope % No. of Effort Peat- Peatland No. of Pool area No. recaps perimeter (% emergent pH sample (mins/ land area (m2) pools (m2) caught (btw (m) steep) vegetation days m2) days) 96.3 4.5 A 3052 276 111 12.9 (5.2) 8.1 (1.0) 6 93 18 0.41 (0.8) (0.10) 91.9 4.3 B 10 729 886 476 3.7 (0.5) 17.2 (0.2) 5 171 19 0.31 (0.2) (0.01) 83.5 4.4 D 13 746 856 385 9.5 (2.6) 10.1 (0.4) 6 213 18 0.17 (0.4) (0.02) 89.2 4.3 E 6523 482 74 9.9 (5.7) 16.5 (0.7) 6 128 6 0.22 (0.7) (0.04) 77.9 4.1 H 4073 448 293 2.4 (0.6) 15.5 (0.6) 6 48 14 0.27 (0.6) (0.02) 85.7 4.3 I 34 676 1619 1020 8.0 (1.9) 6.6 (0.3) 7 207 36 0.11 (0.4) (0.03) 91.2 4.0 J 11 239 618 518 10.7 (2.8) 6.2 (0.5) 5 48 9 0.12 (0.5) (0.03) 92.5 4.9 K 10 724 707 263 7.8 (1.3) 11.0 (0.3) 2 56 2 0.13 (0.3) (0.01) 96.5 4.1 L 4106 473 254 7.5 (3.4) 6.9 (0.7) 5 37 7 0.23 (0.7) (0.03) 81.9 4.3 M 4366 584 59 20.3 (11.0) 6.6 (0.6) 6 143 54 0.38 (0.6) (0.02)

93.8 4.2 N 2613 312 135 2.4 (0.8) 12.3 (0.7) 4 15 4 0.34 (0.7) (0.08) 93.1 4.2 O 7301 503 588 4.5 (1.1) 10.3 (0.5) 6 106 23 0.20 (0.5) (0.03)

63

Table 2.2. Multi-strata models for live recaptures exploring daily survival (Φ), recapture (p), and movement rates (Ψ) of L. hudsonica (with standard errors) for 6 functional groups of peatlands in the Humber River watershed, 2003. C-hat value = 1.8665.

95% CI Parameter Estimate Standard Error Lower Upper Φ (.) 0.874 0.0129 0.846 0.897 p (A) 0.0727 0.0158 0.047 0.110 p (B) 0.0928 0.0255 0.054 0.156 p (H) 0.132 0.0645 0.048 0.314 p (I) 0.0977 0.0196 0.065 0.143 p (M) 0.276 0.0460 0.195 0.374 p (O) 0.213 0.0515 0.129 0.331 Ψ A to B 0.0130 0.0099 0.003 0.056 Ψ A to H 0.00310 0.0030 0.000 0.020 Ψ A to I 0.0101 0.0044 0.004 0.024 Ψ A to M 0.0108 0.0037 0.006 0.021 Ψ A to O 0.00580 0.0023 0.003 0.012 Ψ B to A 0.0149 0.0112 0.003 0.063 Ψ B to H 0.00210 0.0024 0.000 0.019 Ψ B to I 0.00740 0.0030 0.003 0.016 Ψ B to M 0.00770 0.0027 0.004 0.015 Ψ B to O 0.00360 0.0031 0.001 0.019 Ψ H to A 0.0135 0.0141 0.002 0.098 Ψ H to B 0.00800 0.0095 0.001 0.079 Ψ H to I 0.0944 0.0744 0.019 0.365 Ψ H to M 0.0349 0.0257 0.008 0.139 Ψ H to O 0.0145 0.0115 0.003 0.067 Ψ I to A 0.00240 0.0019 0.001 0.011 Ψ I to B 0.00150 0.0012 0.000 0.007 Ψ I to H 0.00540 0.0052 0.001 0.035 Ψ I to M 0.00170 0.0014 0.000 0.009 Ψ I to O 0.00120 0.0011 0.000 0.007 Ψ M to A 0.0146 0.0050 0.007 0.028 Ψ M to B 0.00910 0.0032 0.005 0.018 Ψ M to H 0.0110 0.0055 0.004 0.029 Ψ M to I 0.00990 0.0052 0.004 0.027 Ψ M to O 0.0147 0.0093 0.004 0.050 Ψ O to A 0.00860 0.0033 0.004 0.018 Ψ O to B 0.00470 0.0039 0.001 0.024 Ψ O to H 0.00490 0.0031 0.001 0.017 Ψ O to I 0.00740 0.0043 0.002 0.023 Ψ O to M 0.0160 0.0101 0.005 0.054

64

Table 2.3. Best multi-strata models explaining movement of L. hudsonica between functional groups of peatlands. Daily survival rate (Φ) was constant and daily recapture rate (p) differed between functional groups of peatlands. MATRIX covariates included the ln distance of SCRUB, FOREST, and CUT MATRIX and DISTANCE was the ln closest distance separating functional groups of peatlands. WATER (out) was the mean surface area of water in the functional groups of peatlands that were emigrated out of and EMERGENT (in) was the mean percent cover of emergent vegetation of peatlands that were immigrated into.

QAICc No. Q Model QAICc ∆QAICc weight Par Deviance Φ(.), p(peatland), FOREST, CUT, SCRUB, WATER (out) 1232.11 0.00 0.231 12 373.2 Φ(.), p(peatland), DIST, WATER (out), EMERGENT (in) 1232.46 0.35 0.194 11 375.6 Φ(.), p(peatland), FOREST, CUT, SCRUB, WATER, EMERGENT (in) 1232.61 0.50 0.180 13 371.7 Φ(.), p(peatland), DIST 1233.48 1.36 0.117 9 380.7

65

Table 2.4. Parameter estimates for co-variates of the top 4 multi-strata models explaining movement of L. hudsonica between peatlands, including standard errors and 95% confidence intervals. Link function = logit.

95% confidence interval Standard Model Parameter Beta Lower Upper Error FOREST + CUT + SCRUB + WATER (out) Intercept 9.25 5.17 -0.88 19 FOREST -0.71 0.52 -1.7 0.30 CUT -0.03 0.14 -0.31 0.25 SCRUB 0.34 0.28 -0.20 0.88 WATER (out) -1.46 0.61 -2.7 -0.26 DIST + WATER (OUT) + EMERGENT (in) Intercept 4.91 4.72 -4.3 14 DIST -0.42 0.50 -1.4 0.56 WATER (out) -1.36 0.64 -2.6 -0.11 EMERGENT (in) 0.03 0.08 -0.12 0.19 FOREST + CUT + SCRUB + WATER (out) + EMERGENT (in) Intercept 8.74 5.23 -1.5 19 FOREST -0.65 0.51 -1.6 0.35 CUT -0.04 0.15 -0.33 0.24 SCRUB 0.49 0.30 -0.10 1.1 WATER (out) -1.68 0.64 -2.9 -0.44 EMERGENT (in) 0.12 0.09 -0.060 0.29 DIST Intercept -5.01 0.37 -5.7 -4.3

DIST -0.52 0.46 -1.4 0.39

66

Table 2.5. Parameter estimates and standard errors of the co-variates for the glm with a binomial distribution estimating the influence of DISTANCE and FOREST MATRIX between peatlands on the number of male L. hudsonica moving.

Estimate Std. Error P(>|z|) Intercept -4.31 0.25 < 0.001 Distance 0.00171 0.00078 0.029 Forest matrix -0.00524 0.0019 0.0057 distance*forest matrix 3.87e-06 1.5e-06 0.00081

Table 2.6. Residual deviances for the glm with a binomial distribution estimating the influence of DISTANCE and FOREST MATRIX on the number of male L. hudsonica moving.

Residual deviance Residual df P(>|Chi|) Null 114.6 65 Distance 94.6 64 < 0.001 Forest matrix 93.0 63 0.21 distance*forest matrix 82.9 60 0.0010

67

General Discussion

This study was conducted on two spatial scales; the scale of the peatland and of the

landscape. I expanded our knowledge on the population dynamics of L. hudsonica

through the examination of their movement potential and their relationship with both

peatland and landscape features. I demonstrated that at both scales the area of water (i.e.

breeding sites) was an important factor affecting the spatial structure of the population.

At the fine scale, male L. hudsonica were found to aggregate around large pools while

area of water was inversely related to movement between peatlands at broader scales. I

also demonstrated in addition to simple Euclidean distance, the matrix influences inter-

peatland movement rates but that this relationship is not simple.

This study demonstrates the importance of the process of movement in structuring these

populations at both the fine and broad-scales. Further research is required to link the fine

and broader-scale movement behaviours to regional distribution and long term population

dynamics. Our ability to predict across scales is still elementary, however, there has been

some progress made. Jonsen and Taylor (2000a) have demonstrated that it is possible to

predict broad scale patterns of distribution with fine-scale movement behaviours. They

generated simulation models in which they found that fine-scale movement behaviours

were able to predict landscape-level population distribution patterns, once information about the matrix was included. In addition, Krawchuk and Taylor (2003) demonstrated that local-scale information increased their understanding of broad-scale phenomena.

68

Long-term research is also needed to observe how landscapes, especially clearcuts,

structurally change through time, and how this affects population dynamics and

landscape connectivity. Does regenerating forest have the structural appearance of

scrub? How does the age of the clearcut surrounding a peatland influence the maiden

flight? If tenerals must fly further to access appropriate forest habitat, is there a

consequence for survival?

This study was conducted ca. 30 km east from the Gros Morne National Park (GMNP)

boundary. There is concern among biologists, conservationists, government officials, and

the general public that the commercial timber harvesting that is currently taking place in

the Gros Morne Greater Ecosystem will affect the long-term ecological integrity of the

park by functionally isolating it from the broader landscape. Since 2001, forestry

operators have co-operatively worked to develop science-based solutions to minimize the

effect of harvesting on the connectivity between GMNP and its surroundings. Work on

L. hudsonica adds useful information regarding appropriate scales of observation and the importance of the process of movement to regional animal populations.

69

References

Addicott, J. F., J. M. Aho, M. F. Antolin, D. K. Padilla, J. S. Richardson, and D. A. Soluk. 1987. Ecological neighborhoods: scaling environmental patterns. Oikos 49:340-346.

Alcock, J. 1987a. The effects of experimental manipulation of resources on the behavior of two calopterygid damselflies that exhibit resource-defense polygyny. Canadian Journal of Zoology 65:2475-2482.

Alcock, J. 1987b. Male reproductive tactics in the libellulid dragonfly Paltothemis lineatipes: temporal pertitioning of territories. Behaviour 103:157-173.

Alcock, J. 1990. Oviposition resources, territoriality and male reproductive tactics in the dragonfly Paltothemis lineatipes (Odonata: Libellulidae). Behaviour 113:251- 263.

Allen, C. R., and C. S. Holling. 2002. Cross-scale structure and scale breaks in ecosystems and other complex systems. Ecosystems 5:315-318.

Angelibert, S., and N. Giani. 2003. Dispersal characteristics of three odonate species in a patchy habitat. Ecography 26:13-20.

Baguette, M., S. Petit, and F. Quéva. 2000. Population spatial structure and migration of three butterfly species within the same habitat network: consequences for conservation. Journal of Applied Ecology 37:100-108.

Bean, K., G. P. Jones, and M. J. Caley. 2002. Relationships among distribution, abundance and microhabitat specialisation in a guild of coral reef triggerfish (family Balistidae). Marine Ecology Progress Series 233:263-272.

Bélisle, M., and C. C. St. Clair. 2001. Cumulative effects of barriers on the movements of forest birds. Conservation Ecology 5: http://ns2.resalliance.org/pub/www/Journal/vol5/iss2/art9/.

Betzholtz, P.-E. 2002. Population structure and movement patterns within an isolated and endangered population of the moth Dysauxes ancilla L. (Lepidoptera, Ctenuchidae): implications for conservation. Journal of Insect Conservation 6:57- 66.

Blaustein, L., M. Kiflawi, A. Eitam, M. Mangel, and J. E. Cohen. 2004. Oviposition habitat selection in response to risk of predation in temporary pools: mode of detection and consistency across experimental venue. Oecologia 138:300-305.

Bohonak, A. J. 1999. Dispersal, gene flow, and population structure. The Quarterly Review of Biology 74:21-45.

70

Botero-Garcés, N., and R. Issacs. 2004. Movement of the grape berry moth Endopiza viteana: displacement distance and direction. Physiological Entomology 29:443- 452.

Bowler, D. E., and T. G. Benton. 2005. Causes and consequences of animal dispersal strategies: relating individual behaviour to spatial dynamics. Biological Reviews 80:205-225.

Brown, J. H., and A. Kodric-Brown. 1977. Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58:445-449.

Browne, D. R., J. D. Peles, and G. W. Barrett. 1999. Effects of landscape spatial structure on movement patterns of the hispid cotton rat (Sigmodon hispidus). Landscape Ecology 14:53-65.

Buchwald, R. 1992. Vegetation and dragonfly fauna - characteristics and examples of biocenological field studies. Vegetatio 101:99-107.

Chardon, J. P., F. Adriaensen, and E. Matthysen. 2003. Incorporating landscape elements into a connectivity measure: a case study for the Speckled wood butterfly (Parage aegeria L.). Landscape Ecology 18:561-573.

Colwell, M. A., and S. L. Landrum. 1993. Nonrandom shorebird distribution and fine- scale variation in prey abundance. The Condor 95:94-103.

Conrad, K. F., K. H. Willson, I. F. Harvey, C. J. Thomas, and T. N. Sherratt. 1999. Dispersal characteristics of seven odonate species in an agricultural landscape. Ecography 22:524-531.

Cooch, E., and G. White. 2005. Program MARK: A Gentle Introduction. http://www.phidot.org/software/mark/docs/book/

Corbet, P. S. 1980. Biology of Odonata. Annual Review of Entomology 25:189-217.

Corbet, P. S. 1999. Dragonflies: Behavior and Ecology of Odonata. Cornell University Press, Ithaca.

Cordero, A. 1995. Correlates of male mating success in two natural populations of the damselfly Ischnura graellsii (Odonata: Coenagrionidae). Ecological Entomology 20:213-222.

Courchamp, F., T. Clutton-Brock, and B. Grenfell. 1999. Inverse density dependence and the Allee effect. Trends in Ecology and Evolution 14:405-410.

Crawley, M. 2002. Statistical Computing: An Introduction to Data Analysis using S-Plus. John Wiley and Sons, Ltd., New York, USA.

71

Cressie, N. A. C. 1993. Statistics for Spatial Data, Revised edition. John Wiley & Sons, Inc., New York.

Crist, T. O., D. S. Guertin, J. A. Wiens, and B. T. Milne. 1992. Animal movement in heterogeneous landscapes: an experiment with Eleodes beetles in shortgrass prairie. Functional Ecology 6:536-544.

Cronin, J. T. 2003. Movement and spatial population structure of a prairie planthopper. Ecology 84:1179-1188.

Dalthorp, D. 2004. The generalized linear model for spatial data: assessing the effects of environmental covariates on population density in the field. Entomologia Experimentalis et Applicata 111:117-131.

Dunning, J. B., B. J. Danielson, and H. R. Pulliam. 1992. Ecological processes that affect populations in complex landscapes. Oikos 65:169-175.

Earn, D. J. D., S. A. Levin, and P. Rohani. 2000. Coherence and conservation. Science 290:1360-1364.

Fahrig, L., and G. Merriam. 1985. Habitat patch connectivity and population survival. Ecology 66:1762-1768.

Fahrig, L., and J. E. Paloheimo. 1986. Interpatch dispersal of the cabbage butterfly. Canadian Journal of Zoology 65:616-622.

Flierl, G., D. Grünbaum, S. Levin, and D. Olson. 1999. From individuals to aggregations: the interplay between behavior and physics. Journal of Theoretical Biology 196:397-454.

Forsyth, A., and R. D. Montgomerie. 1987. Alternative reproductive tactics in the territorial damselfly Calopteryx maculata: sneaking by older males. Behavioral Ecology and Sociobiology 21:73-81.

Frankham, R., J. D. Ballou, and D. A. Briscoe. 2003. Introduction to Conservation Genetics. Cambridge University Press, Cambridge, UK.

Gill, F. B. 2000. Ornithology, 2nd edition. W. H. Freeman and Company, New York, USA.

Goodwin, B. J., and L. Fahrig. 2002. Effect of landscape structure on the movement behavior of a specialized goldenrod beetle, Trirhabda borealis. Canadian Journal of Zoology 80:24-35.

72

Grover, K. E., and M. J. Thompson. 1986. Factors influencing spring feeding site selection by elk in the Elkhorn Mountains, Montana. Journal of Wildlife Management 50:466-470.

Gustafson, E. J., and R. H. Gardner. 1996. The effect of landscape heterogeneity on the probability of patch colonization. Ecology 71:94-107.

Haddad, N. M., and K. A. Baum. 1999. An experimental test of corridor effects on butterfly densities. Ecological Applications 9:623-633.

Hale, M. L., P. W. W. Lurz, M. D. F. Shirley, S. Rushton, R. M. Fuller, and K. Wolff. 2001. Impact of landscape management of red squirrel populations. Science 293:2246-2248.

Hamner, W. B., and P. P. Hamner. 2000. Behavior of Antarctic krill (Euphausia superba): schooling, foraging, and antipredatory behavior. Canadian Journal of Fisheries and Aquatic Sciences 57:192-202.

Hanski, I., J. Alho, and A. Moilanen. 2000. Estimating the parameters of survival and migration of individuals in metapopulations. Ecology 81:239-251.

Hanski, I., and D. Simberloff. 1997. The metapopulation approach, its history, conceptual domain, and application to conservation. Pages 5-26 in I. A. Hanski and M. E. Gilpin, editors. Metapopulation Biology: Ecology, Genetics, and Evolution. Academic Press, San Diego, USA.

Hanski, I., and C. D. Thomas. 1994. Metapopulation dynamics and conservation: a spatially explicit model applied to butterflies. Biological Conservation 68:167- 180.

Harrison, S. 1989. Long-distance dispersal and colonization in the Bay checkerspot butterfly, Euphydryas editha bayensis. Ecology 70:1236-1243.

Harrison, S. 1991. Local extinction in a metapopulation context: an empirical evaluation. Biological Journal of the Linnean Society 42:73-88.

Hastings, A., and S. Harrison. 1994. Metapopulation dynamics and genetics. Annual Review of Ecology and Systematics 25:167-188.

Hein, S., J. Gombert, T. Hovestadt, and H.-J. Poethke. 2003. Movement patterns of the bush cricket Platycleis albopunctata in different types of habitat: matrix is not always matrix. Ecological Entomology 28:432-438.

Hilton, D. F. 1987. Odonata of peatlands and marshes in Canada. Memoirs of the Entomological Society of Canada 140:57-63.

73

Hilton, D. F. J. 1984. Reproductive behavior of Leucorrhinia hudsonica (Selys) (Odonata: Libellulidae). Journal of the Kansas Entomological Society 57:580- 590.

Hof, J., and C. H. Flather. 1996. Accounting for connectivity and spatial correlation in the optimal placement of wildlife habitat. Ecological Modelling 88:143-155.

Holder, M. L. 2001. The influence of habitat structure on peatland Odonata at local and landscape spatial scales. M.Sc. Acadia University, Wolfville, Canada.

Holling, C. S. 1992. Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62:447-502.

Holt, R. D. 1990. The microevolutionary consequences of climate change. Trends in Ecology and Evolution 5:311-315.

ImageTool Development Team. 2002. UTHSCSA ImageTool for Windows. The University of Texas Health Science Center in San Antonio, San Antonio.

Ims, R. A. 1995. Movmement patterns related to spatial structures. Pages 85-109 in L. Hansson, L. Farhig, and G. Merriam, editors. Mosaic Landscapes and Ecological Processes. Chapman & Hall, London, UK.

IUCN. 2004. 2004 IUCN Red List of Threatened Species. in.http://redlist.org.

Jackson, J. B. C. 1991. Adaptation and diversity of reef corals. BioScience 41:475-482.

Jepsen, J. U., N. E. Eide, P. Prestrud, and L. B. Jacobsen. 2002. The importance of prey distribution in habitat use by arctic foxes (Alopex lagopus). Canadian Journal of Zoology 80:418-429.

Jetz, W., C. Carbone, J. Fulford, and J. H. Brown. 2004. The scaling of animal space use. Science 306:266-268.

Joly, P., C. Miaud, A. Lehmann, and O. Grolet. 2001. Habitat matrix effects on pond occupancy in newts. Conservation Biology 15:239-248.

Jonsen, I. D., R. S. Bourchier, and J. Roland. 2001. The influence of matrix habitat on Aphthona flea beetle immigration to leafy spurge patches. Oecologia 127:287- 294.

Jonsen, I. D., and L. Fahrig. 1997. Response of generalist and specialist insect herbivores to landscape spatial structure. Landscape Ecology 12:185-197.

74

Jonsen, I. D., and P. D. Taylor. 2000a. Calopteryx damselfly dispersions arising from multiscale responses to landscape structure. Conservation Ecology 4: http://www.consecol.org/vol4/iss2/art4.

Jonsen, I. D., and P. D. Taylor. 2000b. Fine-scale movement behaviors of calopterygid damselflies are influenced by landscape structure: an experimental manipulation. Oikos 88:553-562.

Joyce, K. A., J. M. Holland, and C. P. Doncaster. 1999. Influences of hedgerow intersections and gaps on the movement of carabid beetles. Bulletin of Entomological Research 89:523-531.

Jules, E. S., and P. Shahani. 2003. A broader ecological context to habitat fragmentation: why matrix habitat is more important than we thought. Journal of Vegetation Science 14:459-464.

Kadoya, T., S. Suda, and I. Washitani. 2004. Dragonfly species richness on man-made ponds: effects of pond size and pond age on newly established assemblages. Ecological Research 19:461-467.

Kaluzny, S. P., S. C. Vega, T. P. Cardoso, and A. A. Shelly. 1996. S+ SpatialStats: User's Manual, 1st edition. MathSoft, Inc., Seattle, USA.

Kareiva, P., and U. Wennergren. 1995. Connecting landscape patterns to ecosystem and population processes. Nature 373:299-302.

Kasuya, E., Y. Mashima, and J. Hirokawa. 1987. Reproductive behavior of the dragonfly, Orthetrum japonicum (Odonata: Libellulidae). Journal of Ethology 5:105-113.

Koenig, W. D. 1999. Spatial autocorrelation of ecological phenomena. Trends in Ecology and Evolution 14:22-25.

Koenig, W. D., D. Van Vuren, and P. N. Hooge. 1996. Detectability, philopatry, and the distribution of dispersal distances in vertebrates. Trends in Ecology and Evolution 11:514-517.

Kokko, H., M. P. Harris, and S. Wanless. 2004. Competition for breeding sites and site- dependent population regulation in a highly colonial seabird, the common guillemot Uria aalge. Journal of Animal Ecology 73:367-376.

Kotilar, N. B., and J. A. Wiens. 1990. Multiple scales of patchiness and patch structure: a hierarchical framework for the study of heterogeneity. Oikos 59:253-260.

Krawchuk, M. A., and P. D. Taylor. 2003. Changing importance of habitat structure across multiple scales for three species of insects. Oikos 103:153-161.

75

Kwiatkowski, M. A., and B. K. Sullivan. 2002. Mating system structure and population density in a polygynous lizard, Sauromalus obesus (= ater). Behavioral Ecology 13:201-208. Lancaster, J., B. J. Downes, and P. Reich. 2003. Linking landscape patterns of resource distribution with models of aggregation in ovipositing stream insects. Journal of Animal Ecology 72:969-978.

Lande, R., S. Engen, and B.-E. Sæther. 1999. Spatial scale of population synchrony: environmental correlation versus dispersal and density regulation. The American Naturalist 154:271-281.

Larsen, K. W., and S. Boutin. 1994. Movements, survival, and settlement of red squirrel (Tamiasciurus hudsonicus) offspring. Ecology 75:214-223.

Legendre, P., M. R. T. Dale, M.-J. Fortin, J. Gurevitch, M. Hohm, and D. Myers. 2002. The consequences of spatial structure for the design and analysis of ecological field surveys. Ecography 25:601-615.

Legendre, P., and L. Legendre. 1998. Numerical Ecology, 2nd edition. Elsevier, Amsterdam, Netherlands.

Levin, S. A. 1992. The problem of pattern and scale in ecology. Ecology 73:1943-1967.

Liebhold, A., W. D. Koenig, and O. N. Bjørnstad. 2004. Spatial synchrony in population dynamics. Annual Review of Ecology, Evolution, and Systematics 35:467-490.

Lindstedt, S. L., B. J. Miller, and S. W. Buskirk. 1986. Home range, time, and body size in mammals. Ecology 67:413-418.

Matter, S. F., J. Roland, N. Keyghobadi, and K. Sabourin. 2003. The effects of isolation, habitat area and resources on the abundance, density and movement of the butterfly Parnassius smintheus. The American Midland Naturalist 150:26-36.

Mauremooto, J. R., S. D. Wratten, S. P. Worner, and G. L. A. Fry. 1995. Permeability of hedgerows to predatory carabid beetles. Agriculture, Ecosystems and Environment 52:141-148.

May, R. M. 1994. The effects of spatial scale on ecological questions and answers. Pages 1-17 in P. J. Edwards, R. M. May, and N. R. Webb, editors. Large-scale Ecology and Conservation Biology. Blackwell Scientific Publications, Oxford, UK.

McCarthy, J. 2001. Gap dynamics of forest trees: a review with particular attention to boreal forests. Environmental Review 9:1-59.

76

McIntyre, N. E., and J. A. Wiens. 1999. Interactions between landscape structure and animal behavior: the roles of heterogeneously distributed resources and food deprivation on movement patterns. Landscape Ecology 14:437-447.

McIvor, C. C., and W. E. Odum. 1988. Food, predation risk, and microhabitat selection in a marsh fish assemblage. Ecology 69:1341-1351.

McPherson, A. M. 2003. Between-patch movements of a peatland dragonfly (Leucorrhinia hudsonica), and the influence of landscape structure on its distribution and abundance in western Newfoundland. M.Sc. Acadia University, Wolfville, Canada.

Mech, S. G., and J. G. Hallet. 2001. Evaluating the effectiveness of corridors: a genetic approach. Conservation Biology 15:467-474.

Michiels, N. K., and A. A. Dhondt. 1990. Costs and benefits associated with oviposition site selection in the dragonfly Sympetrum danae (Odonata: Libellulidae). Animal Behaviour 40:668-678.

Moore, A. J. 1989. The behavioral ecology of Libellula luctuosa (Burmeister) (Odonata: Libellulidae): III. Male density, OSR, and male and female mating behavior. Ethology 80:120-136.

Nager, R. G., A. R. Johnson, V. Boy, M. Rendon-Martos, J. Calderon, and F. Cézilly. 1996. Temporal and spatial variation in dispersal in the greater flamingo (Phoenicopterus ruber roseus). Oecologia 107:204-211.

Needham, J. G., M. J. Westfall, Jr., and M. L. May. 2000. Dragonflies of North America, Revised Edition. Scientific Publishers, Inc., Gainesville, USA.

Neubauer, K., and G. Rehfeldt. 1995. Roosting site selection in the damselfly species Calopteryx haemorrhoidalis (Odonata: Calopterygidae). Entomology Generalis 19:291-302.

Nikula, B., J. Sones, D. Stokes, and L. Stokes. 2002. Beginner's Guide to Dragonflies. Little, Brown and Company, Boston, USA.

Opdam, P., and D. Wascher. 2004. Climate change meets habitat fragmentation: linking landscape and biogeographical scale levels in research and conservation. Biological Conservation 117:285-297.

Pebesma, E. J. 2004. Multivariable geostatistics in S: the gstat package. Computers and Geosciences 30:683-691.

Pither, J., and P. D. Taylor. 1998. An experimental assessment of landscape connectivity. Oikos 83:166-174.

77

Plainstow, S., and M. T. Siva-Jothy. 1996. Energetic constraints and male mate-securing tactics in the damselfly Calopteryx splendens xanthostoma (Charpentier). Proceedings of the Royal Society of London Series B Biological Sciences 263:1233-1239.

Purse, B. V., G. W. Hopkins, K. J. Day, and D. J. Thompson. 2003. Dispersal characteristics and management of a rare damselfly. Journal of Applied Ecology 40:716-728.

R Development Core Team. 2004. R: A language and environment for statistical computing. in. R foundation for Statistical Computing, ed., Vienna, Austria.

Ricketts, T. H. 2001. The matrix matters: effective isolation in fragmented landscapes. The American Naturalist 158:87-99.

Rietkerk, M., T. Ouedraogo, L. Kumar, S. Sanou, F. van Langevelde, A. Kiema, J. van de Koppel, J. van Andel, J. Hearne, A. K. Skidmore, N. de Ridder, L. Stroosnijder, and H. H. T. Prins. 2002. Fine-scale spatial distribution of plants and resources on a sandy soil in the Sahel. Plant and Soil 239:69-77.

Rith-Najarian, J. 1996. Ecology and biogeography of Odonata in a northern Minnesota mosaic forest landscape: the impact of anthropogenic disturbance on dragonfly communities in the Mississippi River headwaters region. PhD. University of Colorado, Colorado, USA.

Roland, J., N. Keyghobadi, and S. Fownes. 2000. Alpine Parnassius butterfly dispersal: effects of landscape and population size. Ecology 81:1642-1653.

Roland, J., and P. D. Taylor. 1997. Insect parasitoid species respond to forest structure at different spatial scales. Nature 386:710-713.

Root, R. B., and P. M. Kareiva. 1984. The search for resources by cabbage butterflies (Pieris rapae): ecological consequences and adaptive significance and adaptive significance of Markovian movements in a patchy environment. Ecology 65:147- 165.

Roslin, T. 2000. Dung beetle movements at two spatial scales. Oikos 91:323-335.

Russell, R. E., R. K. Swihart, and Z. Feng. 2003. Population consequences of movement decisions in a patchy landscape. Oikos 103:142-152.

Schtickzelle, N., and M. Baguette. 2003. Behavioural responses to habitat patch boundaries restrict dispersal and generate emigration-patch area relationships in fragmented landscapes. Journal of Animal Ecology 72:533-545.

78

Schwarz, P. A., T. J. Fahey, and C. E. McCulloch. 2003. Factors controlling spatial variation of tree species abundance in a forested landscape. Ecology 84:1862- 1878.

Senft, R. L., M. B. Coughenour, D. W. Bailey, L. R. Rittenhouse, O. E. Sala, and D. M. Swift. 1987. Large herbivore foraging and ecological hierarchies. BioScience 37:789-797.

Simpkin, J. L., H. B. Britten, and P. E. Brussard. 2000. Effects of habitat fragmentation and differing mobility on the population structures of a great basin dragonfly (Sympetrum corruptum) and damselfly (Enallagma carunculatum). Western North American Naturalist 60:320-332.

Sork, V. L., J. Nason, D. R. Campbell, and J. F. Fernandez. 1999. Landscape approaches to historical and contemporary gene flow in plants. Trends in Ecology and Evolution 14:219-224.

St. Clair, C. C., M. Bélisle, A. Desrochers, and S. Hannon. 1998. Winter responses of forest birds to habitat corridors and gaps. Conservation Ecology 2: http://www.consecol.org/vol2/iss2/art13/.

Sutcliffe, O. L., C. D. Thomas, T. J. Yates, and J. N. Greatorex-Davies. 1997. Correlated extinctions, colonizations and population fluctuations in a highly connected ringlet butterfly metapopulation. Oecologia 109:235-241.

Switzer, P. V. 1993. Site fidelity in predictable and unpredictable habitats. Evolutionary Ecology 7:533-555.

Switzer, P. V. 1997a. Factors affecting site fidelity in a territorial animal, Perithemis tenera. Animal Behavior 53:865-877.

Switzer, P. V. 1997b. Past reproductive success affects future habitat selection. Behavioral Ecology and Sociobiology 40:307-312.

Switzer, P. V. 2002. Individual variation in the duration of territory occupation by males of the dragonfly Perithemis tenera (Odonata: Libellula). Annals of the Entomological Society of America 95:628-636.

Taylor, P. D., L. Fahrig, K. Henein, and G. Merriam. 1993. Connectivity is a vital element of landscape structure. Oikos 68:571-573.

Thompson, D. J. 1997. Lifetime reproductive success, weather and fitness in dragonflies. Odonatologica 26:89-94.

Thornhill, R., and J. Alcock. 1983. The Evolution of Insect Mating Systems. Harvard University Press, Cambridge, UK.

79

Trzcinski, M. K., S. J. Walde, and P. D. Taylor. 2003. Colonisation of pitcher plant leaves at several spatial scales. Ecological Entomology 28:482-489.

Tsubaki, Y., and T. Ono. 1986. Competition for territorial sites and alternative mating tactics in the dragonfly, Nannophya pygmaea Rambur (Odonata: Libellulidae). Behaviour 97:234-251.

Tsubaki, Y., and T. Ono. 1987. Effects of age and body size on the male territorial system of the dragonfly, Nannophya pygmaea Rambur (Odonata: Libellulidae). Animal Behaviour 35:518-525.

Tsubaki, Y., and T. Ono. 1995. On the cue for male territorial site selection in the dragonfly Nannophya pygmaea: a field experiment. Journal of Ethology 13:105- 111.

Turchin, P. 1998. Quantitative Analysis of Movement: Measuring and Modeling Population Redistribution in Animals and Plants. Sinauer Associates, Inc., Sunderland, USA.

Turner, M. G. 1989. Landscape ecology: The effect of pattern and process. Annual Review of Ecology and Systematics 20:171-197.

Turner, M. G., V. H. Dale, and R. H. Gardner. 1989. Predicting across scales: theory development and testing. Landscape Ecology 3:245-252.

Ubukata, H. 1981. Survivorship curve and annual fluctuation in the size of emerging population of Cordulia aenea amurensis Selys (Odonata: Corduliidae). Japan Journal of Ecology 31:335-346.

Venables, W. N., and B. D. Ripley. 2000. S Programming. Springer, New York, USA.

Venables, W. N., and B. D. Ripley. 2002. Modern Applied Statistics with S, 4th edition. Springer, New York, USA.

Waage, J. K. 1978. Oviposition duration and egg deposition rates in Calopteryx maculata (P. de Beauvois) (Zygoptera: Calopterygidae). Odonatologica 7:77-88.

Waage, J. K. 1987. Choice and utilization of oviposition sites by female Calopteryx maculata (Odonata: Calopterygidae). Behavioral Ecology and Sociobiology 20:439-446.

Waltz, E. C., and L. L. Wolf. 1988. Alternative mating tactics in male white-faced dragonflies (Leucorrhinia intacta): plasticity of tactical options and consequences for reproductive success. Evolutionary Ecology 2:205-231.

80

Ward, D., and D. Saltz. 1994. Foraging at different spatial scales: Dorcas gazelles foraging for lilies in the Negev Desert. Ecology 75:48-58.

Watts, P. C., J. R. Rouquette, I. J. Saccheri, K. S. J., and D. J. Thompson. 2004. Molecular and ecological evidence for small-scale isolation by distance in an endangered damselfly, Coenagrion mercuriale. Molecular Ecology 13:2931- 2945.

White, G. C., and R. E. Bennetts. 1996. Analysis of frequency count data using the negative binomial distribution. Ecology 77:2549-2557.

White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46:S120-S138.

Wiens, J. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397.

Wiens, J., N. C. Stenseth, B. Van Horne, and R. A. Ims. 1993. Ecological mechanisms and landscape ecology. Oikos 66:369-380.

Wiens, J. A., R. L. Schooley, and R. D. Weeks Jr. 1997. Patchy landscapes and animal movements: do beetles percolate? Oikos 78:257-264.

With, K. 1994. Using fractal analysis to assess how species perceive landscape structure. Landscape Ecology 9:25-36.

With, K. A., and T. O. Crist. 1996. Translating across scales: simulation species distributions as the aggregate response of individuals to heterogeneity. Ecological Modelling 93:125-137.

Yang, L. H. 2000. Effects of body size and plant structure on the movement ability of a predaceous stinkbug, Podisus maculiventris (Heteroptera: Pentatomidae). Oecologia 125:85-90.

Zollner, P. A. 2000. Comparing the landscape level perceptual abilities of forest sciurids in fragmented agricultural landscapes. Landscape Ecology 15:523-533.

81

Appendix

Pe atland B Pe at lan d K Pe atland M 0 0 0 4 00 2 04 00 0 3 3 15 y 0 0 enc 0 20 0 1 Frequ 02 0 1 100 5 0 0 0

0 1020 30 4050 6070 0 50 100 150 200 0 100 200 300 400 500 600 Pool area (m2)

Appendix 1. Frequency of pool sizes in peatlands B, K, and M.

82

100 0 8 id in gr Peatland B Peatland K 0

6 Peatland M ge pools lar of r e t a 0 w 4

of r e v o c

% 0 2 0

0 20406080100

% of large pools in grid

Appendix 2. The relationship between the proportion of large pools (area > 5 m2) in each grid to the total percent cover of water large pools account for at the 64×64 m2 scale in peatlands B, K, and M.

83

Appendix 3. Results from preliminary analysis of distance from pool to the edge of peatland B.

Median distance (m) from the pool to the edge of the peatland 7.2 No. of large pools within 7.2 m 34 No. of large pools beyond 7.2 m 21 No. of individuals found within 7.2 m 107 No. of individuals found beyond 7.2 m 84 No. of individuals found within 7.2 m and on pools greater than 5 m2 57 No. of individuals found within 7.2 m and on pools less than than 5 m2 50 No. of individuals found beyond 7.2 m and on pools greater than 5 m2 36 No. of individuals found beyond 7.2 m and on pools less than than 5 m2 48