The importance of edge effects in determining fish distributions in patchy habitats

Timothy Malcolm Smith

Thesis submitted in fulfilment of the requirements of the degree of Doctor of Philosophy at the University of

August 2009

Department of Zoology Faculty of Science The University of Melbourne Abstract

Boundaries between adjacent habitats can create unique biotic and abiotic conditions, varying species compositions and abundances between the edge and interior of habitats. As habitats become fragmented, the relative amount of edge increases. Understanding the role that habitat edges have in determining species compositions and abundances is fundamental for conservation and management of habitats, particularly those under threat from fragmentation. Seagrass habitats are common nearshore habitats that harbour a rich and diverse faunal assemblage that are under threat worldwide from human disturbance. Human induced fragmentation, and the propensity of seagrass to form naturally patchy landscapes, makes it an ideal system to study the effects of edges on fauna.

Evidence of fish displaying edge effects in seagrass habitats is equivocal. Assessment of fish edge effects was done by sampling seven positions within seagrass habitats at fine spatial scales. Strong, consistent patterns in fish distributions demonstrated clear edge effects both within and alongside seagrass at these sites. The total number of fish sampled was greater at the seaward seagrass edge than the seagrass middle, but there was little difference between the seagrass middle and the shoreward seagrass edge. Four individual fish species showed preferences for the seagrass edges. Further investigation revealed that patch size could influence the magnitude of edge effects in seagrass beds. Fish were sampled in ten variously sized seagrass patches in three positions within each patch. Two species showed variations in edge effects across patches which could be attributed to the area of the patch. Changes in patch size can influence the magnitude of edge effects that species display, suggesting that patch area effects (fish density varying with patch size) could be caused by edge effects.

Food availability and predation are mechanisms commonly used to explain edge effect patterns. Gut analysis was done on nigra sampled at the edge and middle of patches to determine if prey consumption varied between positions, and explain S. nigra distribution. There was little difference in prey consumed by S. nigra at the edge and middle of patches, suggesting that food was unlikely to be causing S. nigra edge

2 effects, or that the influence of prey distribution was being masked by other factors such as seagrass structure. Predator abundances and foraging efficiency may vary at the edge and middle of patches, and consequently influence the distribution of prey fish within patches. Underwater videos were placed at four positions within seagrass habitats to assess predator distributions. Predatory Australian salmon, Arripis spp., spend more time over adjacent sand than other positions, while small potential prey species (King George whiting, Sillaginodes punctata, recruits) appear to prefer the middle of seagrass patches, possibly to avoid encounters with salmon. To test if the predator-prey distributions reflected actual predation pressure, a tethering experiment was done to determine if predation was causing edge effects in small fishes. King George whiting recruits and (Stigmatopora spp.) were tethered at each of the four positions at different depths. Survival time of whiting recruits was greater in the middle of shallow seagrass patches than other positions. Few pipefish were preyed upon, and survival time was lower over sand adjacent to seagrass than at the seagrass edge or middle. Video footage revealed that salmon was the dominant predator of both whiting recruits and pipefish. The distribution of predators and associated predation can explain edge effects for some species (whiting) but other mechanisms, or a combination of mechanisms, are determining edge effects for other species (pipefish).

Edge effects were common amongst fish species in seagrass habitats, and included permanent, temporary and predatory species. Patch size was found to influence the extent of the edge effect. There was little evidence to support prey consumption as an underlying mechanism causing higher fish abundances at the interior or edge of patches, however there was evidence that predation could be causing edge effects. Changes in fish distributions within seagrass patches due to patch size and predation when seagrass undergoes fragmentation need to be considered by not only ecologists, but also by managers in the development of plans for seagrass conservation. Future studies should investigate the relative contribution of different edge characteristics in determining the degree of seagrass edge effects.

3 Declaration

This is to certify that:

i. the thesis comprises only my original work towards the PhD, ii. due acknowledgment has been made in the text to all the material used, iii. the thesis is less than 100,000 words in length, exclusive of tables, maps, bibliographies and appendices.

Signature______Date______

4 Acknowledgments Firstly I would like to thank my supervisors Jeremy Hindell, Greg Jenkins, Mick Keough and Rod Connolly, who gave me the opportunity to undertake and complete this thesis. Their advice, patience, support and feedback over the years has been invaluable and without them I surely would never have been able to complete this journey. Thank you to everyone who has helped me in the field over the years, particularly sampling during cold, wet nights. Rod, Pete, Dave, Fabian, Hannah, Fi, Jess, Neils, Ted, Claire, Dad and Susi your help in the field will always be remembered. Thanks heaps Jules and Neil for reading over drafts of this thesis. To everyone at the Victorian Marine Science Consortium, Corey, Isla, Pete, Fi, Dave, Fabian, Lauren, Jay, Kathryn, Jo, Justin and particularly Rod and Liz who were always there to help and provided a great working atmosphere. To everyone that has helped me from the Marine and Freshwater Fisheries Research Institute, particularly Paul, Brent, Sean, Dave, Kade, Simon, Dave, Neil and Vicki your time and effort is greatly appreciated. Dave, Fabian, Camilla and Fi have been great housemates to talk over the trials and tribulations of marine science PhD’s and Paul and Ako gave me a bed when I couldn’t find anywhere else for which I am most thankful.

I received financial support from Victorian branch of the Australian Marine Science Association for funding my trip to New Zealand to present at a conference, the Australian Research Council who funded the project and the University of Melbourne who provided me with a scholarship so that I could complete my work.

To all my mates who have continually harassed and encouraged me, I have finally finished, thanks a lot boys!! Thanks Susi for all you support, encouragement and kindness throughout my time studying, without it things would have been a whole lot tougher. Tals, thanks for being there to help me finish this thesis. To Mum, Dad, Claire and the rest of my family, thanks for all the support, money, effort and encouragement, without it I might never have got to where I am today.

5 6 “One Fish, Two Fish Red Fish, Blue Fish” Dr Seuss

7 Contents

List of Tables...... 10

List of Figures...... 12

1. General Introduction ...... 15

Thesis Outline ...... 20

2. Edge effects on fish associated with seagrass and sand patches ...... 22

Introduction...... 23

Methods ...... 25

Results ...... 29

Discussion...... 41

3. Seagrass patch size affects fish responses to edges ...... 45

Introduction...... 46

Methods ...... 48

Results ...... 51

Discussion...... 56

4. Fine-scale spatial and temporal variations in diets of the pipefish within seagrass patches in Bay, , ...... 61

Introduction...... 62

Methods ...... 64

Results ...... 65

Discussion...... 77

5. Edge effects in patchy seagrass landscapes: the role of predation in determining fish distribution...... 81

Introduction...... 82

Methods ...... 84

Results ...... 88

8 Discussion...... 97

6. General Discussion ...... 102

7. Bibliography ...... 108

9 List of Tables

Table 1. Area and average seagrass variables (SE) in each patch at each site...... 26 Table 2. ANOVA results for seagrass biomass, length, leaf density and epiphyte biomass. Significant results in bold (p < 0.05)...... 30 Table 3. Total abundances of fish sampled at each position at all sites and times...... 34 Table 4. ANOVA results for fish density and species richness. Significant results in bold (p < 0.05)...... 35 Table 5. Planned comparison results for fish density and species richness at each site pooled across time of day. (*= < 0.05, **= < 0.01, ***= < 0.001, > < indicate position with the greatest density) ...... 36 Table 6. ANOVA results for individual fish species. Significant results in bold (p < 0.05)...... 37 Table 7. Planned comparison results for individual fish species within each site for Stigmatopora nigra and S. argus (pooled across time of day) and time of day for Nesogobius maccullochi and Cristiceps australis (pooled across site). ( *= < 0.05, ***= < 0.01, ***= < 0.001, > < indicate position with the greatest density)...... 40 Table 8. Landscape characteristics of each patch sampled...... 49 Table 9. Repeated measures analysis between the edge and middle for species richness, Stigmatopora nigra and Nesogobius maccullochi. Significant results in bold (p < 0.05)...... 52 Table 10. Repeated measures analysis between the edge and 2 m for species richness, Stigmatopora nigra and Nesogobius maccullochi. Significant results in bold (p < 0.05)...... 53 Table 11. Mean abundance of each prey type per gut (±1 SE) from Stigmatopora nigra guts and frequency of occurrence. Data pooled across position, season and time of day...... 66 Table 12. Results of repeated measure analysis of variance, assessing Stigmatopora nigra length, total prey abundance, harpacticoid and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05)...... 67 Table 13. Results of repeated measure analysis of variance, assessing male Stigmatopora nigra length, total prey abundance, harpacticoid copepods and

10 planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05)...... 69 Table 14. Results of repeated measure analysis of variance, assessing female Stigmatopora nigra length, total prey abundance, harpacticoid copepods and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05)...... 70 Table 15. Results of repeated measure analysis of variance, assessing juvenile Stigmatopora nigra length, total prey abundance, harpacticoid copepods and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05)...... 71 Table 16. Results of repeated measures analysis of variance comparing seagrass density among positions and season. Significant results in bold (p < 0.05)...... 74 Table 17. Analysis of variance comparing the time spent and MaxN in each position for each species sampled with videos. Significant values in bold (p < 0.05)...... 90 Table 18. Analysis of variance comparing survival times of King George whiting recruits and pipefish at different positions and seagrass bands. Significant values in bold (p < 0.05)...... 95

11 List of Figures

Figure 1. Location of study sites in Port Phillip Bay. Insert: Location of Port Phillip Bay within Australia...... 25 Figure 2. Positions sampled within each seagrass patch. At each patch SG Mid was equal to half the distance between SG-E Sea and SG-E Shore, which was the same distance UV Sea and Shore were from the seagrass edge. Dotted lines represent all planned comparisons except between seagrass and unvegetated positions...... 27 Figure 3. Average seagrass biomass, length, leaf density and epiphyte biomass at each site and position (± 1 SE)...... 31 Figure 4. Average depth at each position within each site (± 1SE)...... 32 Figure 5. Fish density (fish m-2) and species richness (species m-2) within each position, at each site during the day and night (± 1 SE)...... 36 Figure 6. Average density at each site of Stigmatopora nigra and S. argus at each position pooled over diel periods (no significant interaction with time of day), and Nesogobius maccullochi and Cristiceps australis at each position during each diel period (±1 SE)...... 38 Figure 7. Schematic representation of changes in strength and direction of edge effects as patch size increases if edge effects and patch area interact. Diagrams on the left show density differences at the edge and middle as patch area increases, and the consequent changes in edge effect strength and direction on the right...... 47 Figure 8. Relationship between mean species richness and seagrass density across patches at the seagrass edge and 2 m (pooled)...... 54 Figure 9. Relationship between the difference in density at the edge and interior (edge –

interior), and seagrass patch area (log10) for (a) the pipefish Stigmatopora nigra and (b) the goby Nesogobius maccullochi. Data pooled over season and time of day...... 55 Figure 10. Average Stigmatopora nigra density at the edge, 2 m and middle during each season and time of day (±1 SE)...... 56 Figure 11. Mean (± 1 SE) Stigmatopora nigra length during (a) autumn and spring, and (b) at the edge and middle of seagrass patches, and (c) mean (± 1 SE) S. nigra harpacticoid gut abundance during autumn and spring...... 68

12 Figure 12. Mean (± 1 SE) male and female Stigmatopora nigra length, and female harpacticoid copepod prey abundance during autumn and spring pooling across positions, patches and time of day...... 73 Figure 13. Relationship between mean(a) Stigmatopora nigra lengths across sampling days and mean harpacticoid copepod abundance, (b) female S. nigra length and harpacticoid copepod abundance, (c) female length and seagrass density, (d) male S. nigra lengths and total prey abundance and (e) male length and planktonic copepod abundance...... 75 Figure 14. Relationship between mean seagrass density and mean Stigmatopora nigra (a) harpacticoid and (b) planktonic copepod prey, and (c) male S. nigra planktonic copepod prey...... 76 Figure 15. Mean (± 1 SE) TiV in each position for Australian salmon, smooth toadfish, small schooling fish and King George whiting adults, TiV at each position on each sampling day for King George whiting recruits, and mean MaxN (± 1 SE) at each position for small schooling fish and smooth toadfish...... 91 Figure 16. Relationship between each of TiV and MaxN and water depth for Australian salmon and smooth toadfish, and between TiV and seagrass density at the middle and edge of patches for Australian salmon and small schooling fish..... 92 Figure 17. Mean (± 1 SE) TiV and MaxN in each position during night samples for small schooling fish and smooth toadfish...... 93 Figure 18. Mean (± 1 SE) survival time of King George whiting recruits across positions and seagrass bands, and pipefish across positions...... 96 Figure 19. Relationship between the probability of King George whiting recruits being eaten after 30 min and King George whiting length...... 97

13 Publication Details

Smith TM, Hindell JS, Jenkins GP, Connolly RM (2008) Edge effects on fish associated with seagrass and sand patches. Mar Ecol Prog Ser, 359: 203-213

Smith TM, Hindell JS, Jenkins GP, Connolly RM (2010) Seagrass patch size affects fish response to edges. J Anim Ecol, 79: 275-281

14 1. General Introduction

Landscape attributes play a major role in determining the characteristics of biological communities. Proximity to alternative habitats, patch size and within patch location can all influence the fauna and flora that inhabit a patch (Bender et al. 1998, Fahrig 2003). Edge effects relate to the influence that a patch edge can have in determining species composition and processes within a patch and are a key component in understanding impacts of habitat fragmentation (Ries et al. 2004, Fletcher et al. 2007). Edges are a conspicuous feature of landscapes, providing a boundary between adjacent habitats of different qualities and are often ecologically distinct from patch interiors (Ries et al. 2004). Habitat boundaries are ecologically important, changing physical variables, species interactions and flows which can affect species abundances (Murcia 1995, Fagan et al. 1999, Cadenasso et al. 2003, Ries et al. 2004). Edges can have a positive, negative, or no effect on species, depending on the species in question and the type of habitat edge (Ries & Sisk 2004). The importance of edges in the distribution and abundance of species has been know since Leopold (1933) used the term “edge effects” to describe an increase in game species in patchy landscapes. Historically, habitat edges were thought to increase species diversity and abundance; however, edges have recently been associated with the invasion of exotic species, decline in songbirds and quality core habitat for specialist species (Ries et al. 2004). The possibility that edges are an undesirable landscape feature has generated a plethora of studies focusing on the relationship between edges and birds (Paton 1994), mammals (Kremsater & Bunnell 1992), plants (Laurance et al. 1998) and insects (Barbosa & Marquet 2002), with varying results. As habitat loss and fragmentation continue to increase as a result of human disturbance, landscape ecology, and understanding edge effects, become critical for habitats and species conservation and management (Ries et al. 2004, Fletcher et al. 2007).

An edge can be defined as the boundary between two distinct patches, and is therefore dependent on the definition of patch within a landscape context (Ries et al. 2004). What constitutes a patch is determined by the research question being asked and can be distinguished by composition, structure or function at a given scale (Cadenasso et al.

15 2003). Edge characteristics are determined by patch quality, resulting in varying responses to edges reported in the literature (Ries et al. 2004). Edge characteristics such as size, contrast, shape and geometry can vary according to patch characteristics, and determine the importance of edges for species and ecological flows (Strayer et al. 2003). Such variation of edge characteristics can to inconsistent results among studies, and need to become consistent when comparing studies to gather meaningful information (Murcia 1995, Strayer et al. 2003, Ries et al. 2004).

A range of mechanisms have been described to explain edge effects. Ries et al. (2004) identified four key mechanisms that affect species abundances across habitat edges, ecological flows such as materials, energy, organisms and information (Cadenasso et al. 2003), access to resources, resource mapping and species interactions. Ecological flows can be amplified, attenuated or reflected by habitat edges depending on the permeability of the edge (Strayer et al. 2003), which is largely governed by the structure of the edge (Cadenasso et al. 2003). Access directly affects ecological processes. Edges provide access to resources that are spatially separated by a boundary, maximising species ability to use resources from both patches (Ries et al. 2004). Resource mapping influences edge effects when the distribution of an organism’s resources change at the edge, causing a change in the species distribution (Ries et al. 2004). Species interactions at habitat edges can be altered by creating novel encounters, altering movement patterns, mortality rates and influx of species from other habitats (Fagan et al. 1999). Using these mechanisms Ries et al. (2004) developed a model that predicts positive, negative or neutral responses of species to a habitat edge depending on resource allocation within each patch. However, other ecological factors such as edge orientation, habitat contrast, multiple edges, fragmentation and temporal scales can alter the model (Ries et al. 2004).

Seagrasses are a group of completely aquatic angiosperms comprised of approximately 60 species from three families that can be structurally diverse (Orth et al. 2006a) and grow at depths of up to 70 m (Gillanders 2007). Seagrass ecosystems are a prominent nearshore habitat world wide, covering between 0.1 and 0.2 % of coastal oceans (Duarte

16 2002). Seagrass habitats are highly productive, rivalling the most productive habitats on earth, with net production thought to be around 0.6 x 1015 g C yr-1 (Duarte & Chiscano 1999), much of which is exported into other ecosystems (Duarte 2002). play a variety of important functions in coastal ecosystems such as sediment stabilisation, nutrient trapping and cycling, and wave attenuation amongst others (Duarte 2002, Orth et al. 2006a, Gillanders 2007). The distribution and abundance of seagrass habitats along coasts is governed by light and sediment conditions (Duarte 2002), making seagrass habitats susceptible to natural disturbance such as wave energy, turbidity and algal growth (Duarte 2002, Orth et al. 2006a, Gillanders 2007). Recently, seagrass habitats have come under threat from a range of anthropogenic sources including eutrophication, climate change, dredging and shoreline development (Duarte 2002, Orth et al. 2006a), leading to increased habitat loss and fragmentation, creating increased edge habitat (Horinouchi 2007b).

Seagrass habitats support rich faunal assemblages, often greater than nearby habitats such as unvegetated sand (Ferrell & Bell 1991, Jenkins et al. 1997), algal habitats (Sogard & Able 1993), and and saltmarshes (Bloomfield & Gillanders 2005). Seagrass is home to a range of permanent and temporary invertebrate and fish species (Jackson et al. 2001). Permanent species spend their entire life history within seagrass habitats and are often small and cryptic, although they may spend some time within other habitats (Jackson et al. 2001). In contrast, temporary residents spend only part of their life history within seagrass habitats, often recruiting to seagrass from the plankton after being spawned offshore, before moving into other habitats as adults or sub adults (Jackson et al. 2001). Temporary residents have given rise to the concept that seagrass habitats are a nursery for many fauna (Jackson et al. 2001). Seagrass provides fauna with a variety of benefits such as a refuge from predators, a spawning ground, and food sources such as detritus, algae and seagrass itself (Jackson et al. 2001, Horinouchi 2007b). Fauna within seagrass patches can be influenced by seagrass density and biomass (Horinouchi 2007b), water depth (Jackson et al. 2002), organic matter (Hovel et al. 2002) and sediment particle size (Berkenbusch & Rowden 2007).

17 Landscape ecology has traditionally been restricted to terrestrial studies, however, it is equally applicable to marine landscapes (Robbins & Bell 1994) and recently there has been a increase in the number of marine studies (Hinchey et al. 2008). Seagrass beds form naturally patchy habitats, being comprised of patches of different sizes and shapes, making it an ideal habitat to study landscape ecology (Boström et al. 2006). Seagrass edges provide a sharp contrast from the surrounding sand matrix and are generally easily definable. The change in structure at the seagrass edge can alter physical properties at the edge, including water currents (Peterson et al. 2004), light intensity (Carruthers & Walker 1997) and sedimentation (Scoffin 1970), which may have an impact on faunal distribution by affecting recruitment (Tanner 2003), food supply (Walters & Bell 1986) or burrowing ability (Nel et al. 2001).

Few studies have directly assessed edge effects in seagrass habitats (Connolly & Hindell 2006). Many studies have investigated faunal differences in relation to patch size and assumed increasing edge to area ratio demonstrated an edge effect. Such studies, however, do not demonstrate faunal differences between within patch locations, merely that different sized patches contain differing faunal assemblages (Connolly & Hindell 2006, Jelbart et al. 2007). Studies that have directly investigated edge effects in seagrass have generally focused on invertebrates, with few studies on fish, due to the difficulty in sampling fish at spatially appropriate levels.

Results of invertebrate studies comparing seagrass edges to interior have produced varied and conflicting responses, showing increases (bivalves, Bologna 2006; tanaids, Tanner 2005; nematodes, Bologna 2006), decreases (decapods, Bologna 2006; gastropods, Bologna & Heck 2002; polychaetes, Bell et al. 2001) and no difference (decapods, Sanchez-Jerez et al. 1999b, Bologna & Heck 2002; gastropods, Sanchez- Jerez et al. 1999b, Bologna 2006; isopods, Sanchez-Jerez et al. 1999b, Bologna 2006) in densities at the seagrass edge. Fish responses to seagrass edges show similar inconsistent patterns. Overall fish density and species richness show little response to edge effects (Hovel et al. 2002, Uhrin & Holmquist 2003, Johnson & Heck 2006, Jelbart et al. 2007), yet certain individual species have been shown to respond to edges. Jelbart et al. (2007) showed that the pipefish carinirostris were sampled in

18 greater numbers less than 4 metres from the seagrass edge than the seagrass interior (6- 10 m from edge), while Heithaus et al. (2006) used tracking devices to conclude that the top predator Galeocerdo cuvier preferred the seagrass edge over the seagrass interior and deep water habitats. Hovel et al. (2002) found that the pinfish Lagodon rhomboides was negatively correlated with edge habitats, although this pattern was not consistent over time. The vast majority of fish species, however, show no response to the seagrass edge (Connolly & Hindell 2006).

Lack of patterns associated with seagrass edges in fish populations, while possibly reflecting real patterns, may be a result of sampling techniques (Connolly & Hindell 2006). Sampling techniques to detect edge effects in fish need to be both effective, and at appropriate scales for the target species, which becomes problematic when targeting large mobile species. Uhrin & Holmquist (2003), Johnson & Heck (2006) and Hovel et al. (2002) used small drop samplers, which, while suitable for small species and detecting small-scale spatial patterns, are not effective at catching highly mobile species. Conversely, Jelbart et al. (2007) used a seine net, that allowed for the capture of larger fish but potentially missed small-scale spatial differences in fish abundances. Appropriate methods for sampling across seagrass habitats need to be designed and tested for future assessments of fish response to edge effects.

Failure to detect fish assemblage responses to seagrass edges may be caused by fish response to environmental variables such as depth, current flow and seagrass structure (Orth et al. 1984, Bell & Westoby 1986). Therefore, within patch variations in environmental variables need to be given consideration, as they have the potential to mask patterns associated with seagrass edges (Boström et al. 2006, Connolly & Hindell 2006).

Although responses to seagrass edges are unclear, a variety of models have been suggested to explain within patch variations in seagrass faunal assemblages. Predation and food availability are the most commonly cited models, although others such as larval supply, encounter rates and reproductive success have been suggested (Connolly & Hindell 2006 and references therein), yet few of these models have been sufficiently

19 tested. Only Bologna & Heck (1999), who used tethering experiments to demonstrate both increased predation and growth rates on the scallop Argopecten irradians at the seagrass edge, have used manipulative experiments to directly assess models to explain edge effects. Experimental methods such as tagging, tethering and the deployment of artificial seagrass units (ASUs) have, however, been used to assess predation and growth in relation to patch size for juvenile cod (Laurel et al. 2003), A. irradians (Irlandi 1997, Irlandi et al. 1999), and blue crab Callinectes sapidus (Hovel & Lipcius 2002, Hovel & Fonseca 2005). Further investigation using these experimental methods is required to evaluate the influence and, if any, cause of edge effects in seagrass fish populations.

Thesis Outline The aim of this thesis was to investigate if edge effects contribute to distribution patterns of fish species within seagrass habitats in Port Phillip Bay, Victoria, and, examine possible mechanisms including, patch size, food availability and predation that are causing any edge effects. In Chapter 2 I identify edge effects for seagrass fishes sampled at different positions during both the day and night at three sites. Using spatially-explicit sampling methods, fish densities were found to be generally higher at seagrass and sand edges than the interior of seagrass or sand patches regardless of seagrass structure or depth. Chapter 3 investigates factors that may influence edge effects in seagrass with particular focus on the relationship between edge effects and patch area. The magnitude of edge effects depended on the size of seagrass patches for two species but not for others. Chapters 4 and 5 investigated mechanisms that contribute to the patterns observed in chapters 2 and 3. The role of prey availability in determining edge effects was investigated in Chapter 4. Gut analysis was done on a common species of seagrass fish, Stigmatopora nigra, and found prey differences related to seagrass density. There was however little difference in prey consumption between the edge and middle of seagrass patches. The role that predation plays in determining seagrass edge effects was assessed in Chapter 5. Video sampling techniques were used to assess edge related distribution of predators and other fast moving fishes in seagrass habitats. Subsequently, a tethering experiment was used to determine if predator patterns influence the distribution of prey species. Chapter 6 summarises and discusses

20 my results in light of current literature and discusses the implications for seagrass management and conservation.

21 2. Edge effects on fish associated with seagrass and sand patches

Published: Smith TM, Hindell JS, Jenkins GP, Connolly RM (2008) Mar Ecol Prog Ser, 359: 203-213

ABSTRACT Seagrass beds form naturally patchy habitats with large areas of seagrass-sand interface, or edges. Fish were sampled at three sites in the temperate waters of Victoria, Australia, using small (0.5 m wide) push nets at seven positions: unvegetated sand distant from the patch on the seaward side, the sand edge adjacent to the seagrass on the seaward side, the seagrass edge on the seaward side, the middle of the seagrass patch, the seagrass edge on the shore side of the patch, the sand adjacent to the seagrass on the shoreward side of the patch and unvegetated sand distant from the seagrass on the shoreward side. Samples were taken during the day and night, and seagrass variables collected to describe structural complexity. As expected, more fish were caught in seagrass than over sand. Within seagrass, I found strong and consistent patterns at edges. Regardless of site, the total number of fish sampled was greater at the seaward seagrass edge (484 fish) than the seagrass middle (231), but there was little difference between the seagrass middle and the shoreward seagrass edge (297). Two species of pipefish, (193) and S. nigra (160), were much more abundant at the seaward seagrass edge than the seagrass middle at all sites (54, 46 respectively). The goby Nesogobius maccullochi showed a very different pattern. It was more abundant at the shoreward seagrass edge (127) than over the seagrass middle (31) at all sites, and tended to be more abundant over sand at the edge of seagrass patches than any other sand positions. The weedfish, Cristiceps australis, was significantly more abundant at the seaward seagrass edge (26) than the middle (11) but only at night. Consistent patterns in fish distributions demonstrate clear edge effects both within and alongside seagrass at these sites in south eastern Australia.

Keywords: seagrass, fish, edge effects, diel cycles, Heterozostera nigricaulis, unvegetated habitats, seagrass structure

22 Introduction Landscape attributes play a major role in determining the structure of biological communities (Turner 1989). Proximity, size and within-patch location can influence species diversity, density and interactions within a patch (Fahrig 2003). Edge effects relate to the influence that a patch edge can have in determining species composition and processes within a patch. Edge effects can be considered as, or to influence, patterns in biological and physical parameters such as species richness, predation, food availability, disturbance, temperature and moisture (Murcia 1995). Although well studied in terrestrial habitats (e.g. review by Fagan et al. 1999), the study of habitat edges in marine environments is in its infancy.

Seagrass is a prominent habitat in nearshore locations worldwide, harbouring a rich and diverse assemblage of fauna (Jackson et al. 2001). Increased fragmentation due to human activity is assumed to be a threat to seagrass fauna (Boström et al. 2006). Fragmentation increases the amount of edge associated with seagrass patches and, in conjunction with the propensity of natural seagrass to form patchy mosaics, makes it an ideal habitat to study edge effects (Boström et al. 2006). Many studies have investigated patch size and assumed that changes in fish assemblages were related to edge effects because edge to area ratio increases with decreasing patch size (reviewed by Bell et al. 2001). However, these studies do not actually demonstrate changes in faunal assemblages across patch edges, instead they show that different sized patches contain different faunal assemblages (Connolly & Hindell 2006, Jelbart et al. 2006). Few studies have directly assessed edge effects in seagrass habitats.

Studies directly investigating edge effects in seagrass have mostly focused on invertebrates, with fewer studies on fish because of the difficulty in sampling at appropriate spatial scales (Connolly & Hindell 2006). Studies that have investigated fish patterns at seagrass edges have reported variable responses, and in most cases no response at all (Connolly & Hindell 2006). Fish density and species richness often vary little across patch edges (Hovel et al. 2002, Uhrin & Holmquist 2003, Jelbart et al. 2006), although abundances of individual species may change significantly across patch edges. Jelbart et al. (2006) found that pipefish (Urocampus carinirostris) were more

23 abundant at the edge of seagrass (4 m at patch edge) than the interior (6 - 10 m from edge). Conversely, Hovel et al. (2002) found that abundances of pinfish (Lagodon rhomboides) were lower at seagrass edges, although not at all sampling times. The variability in responses shown by fish to seagrass edges may be, in part, a function of the scale of sampling fish and/or changes in seagrass structure at seagrass edges, causing increases or decreases in fish density.

Fish associated with seagrass have been reported from unvegetated sand adjacent to seagrass patches, which may provide a feeding area in close proximity to shelter (Ferrell & Bell 1991). Ferrell & Bell (1991) found that the abundance and diversity of fish over sand within 10 m of seagrass was greater than 100 m from seagrass, and was often more similar to seagrass itself. In general, however, the extent of utilisation of adjacent unvegetated habitats is poorly known. Within seagrass habitat, depth is known to influence fish assemblages (Anderson 2003, Jackson et al. 2006b). In shallow marine and estuarine waters where depth increases with distance from shore, it is therefore likely that seagrass and sand edges will support different fish assemblages in a shoreward and seaward direction because of the changes in depth.

Variation in fish abundances over diel cycles is well known in seagrass (Edgar & Shaw 1995a, Jackson et al. 2006b) and in other habitats, including coral reefs (Nagelkerken et al. 2000), and unvegetated flats (Gibson et al. 1996). However, edge related patterns in fish assemblage structure related to diel cycles in seagrass habitats has only recently been investigated (Jackson et al. 2006b). Diel changes influence densities of seagrass epifauna (Sanchez-Jerez et al. 1999a) and shallow water predators (Gibson et al. 1996), that in turn may influence within-patch distributions of fish. Diel variability in fish assemblages across patch edges can therefore be predicted, but need further study.

In assessing whether fish assemblages vary across patch edges in ways that are consistent with edge effects this study aims to: (1) quantify how fish assemblages change across seagrass edges, including adjacent sand habitats, (2) determine whether within-patch variability in fish assemblage structure changes with diel periods, patch

24 edge location, and water depth, and (3) determine the extent to which observed patterns vary spatially (among sites).

Methods Study sites Sampling was done in Port Phillip Bay, between October 2005 and January 2006. Port Phillip Bay is a large, semi-enclosed bay that is predominantly marine with a small (1 m) tidal range. Three sites were chosen within Port Phillip Bay, Blairgowrie in the southeast (Site 1), and Grassy Point (Site 2) and Indented Head (Site 3) on the central- western coast of the bay (Figure 1). Samples were taken over a 16 week period where abundances of fish could be influenced by ontogeny, behaviour and turnover. Such temporal variation between replicates adds rigour to our results and increases the generality of our findings.

Figure 1. Location of study sites in Port Phillip Bay. Insert: Location of Port Phillip Bay within Australia.

25 Sites are protected from the predominant south-westerly winds and support patches of the seagrass Heterozostera nigricaulis, running parallel to the shore in depths less than 1.5 m, interspersed by tens of metres of unvegetated sand.

At each site, four seagrass patches were chosen for sampling that where > 50 m apart, > 10 m in diameter and showed distinct seagrass/sand boundaries. Patches ranged in size, shape and seagrass structure (Table 1).

Table 1. Area and average seagrass variables (SE) in each patch at each site.

Area Biomass Leaf Density Epiphyte biomass Site Patch Length (mm) (m-2) (g. m-2) (m-2) (g. m-2) Site 1 1 3098 174 (11) 11980 (625) 151 (5) 28 (1.9) 2 461 66 (2) 8341 (215) 85 (>0.1) 4 (0.4) 3 269 141 (4) 10643 (80.) 145 (2) 13 (0.8) 4 5934 125 (6) 9793 (240) 112 (2) 3 (0.4)

Site 2 1 2851 165 (8) 21561 (937) 141 (4) 3 (0.2) 2 5297 139 (9) 18453 (857) 106 (3) 7 (0.8) 3 2740 139 (5) 15931 (430) 133 (3) 41 (3.0) 4 747 224 (1) 20333 (429) 195 (4) 11 (0.9)

Site 3 1 2178 177 (17) 17637 (1227) 128 (7) 12 (1.6) 2 402 59 (4) 9436 (472) 79 (2) 12 (1.0) 3 2238 131 (8) 16680 (593) 112 (4) 35 (2.5) 4 11518 112 (3) 14947 (170) 120 (4) 48 (3.4)

26 Sampling Fish Fish were sampled in the 4 patches at the 3 sites during the day and night to assess fish patterns and diel periods. Seven positions were sampled at each patch (Figure 2): 1. The unvegetated sand in the seaward region at the equivalent distance from the seagrass patch edge, as the middle of the seagrass patch is to the seagrass edge (UV Sea). 2. unvegetated sand within 1 m of the seagrass/sand interface in the seaward region (UV-E Sea) 3. seagrass within 1 m of the seagrass/sand interface in the seaward region (SG-E Sea); 4. the seagrass middle (SG Mid); 5. seagrass within 1 m of the seagrass/sand interface in the shoreward region (SG- E Shore); 6. unvegetated sand within 1 m of the seagrass/sand interface in the shoreward region (UV-E Shore); and, 7. unvegetated sand in the shoreward region at the equivalent distance from the patch edge, as the middle of the seagrass patch is to the seagrass edge (UV Shore).

Seaward Shoreward

UV UV-E SG-E SG Mid SG-E UV-E UV Sea Sea Sea Shore Shore Shore r

Figure 2. Positions sampled within each seagrass patch. At each patch SG Mid was equal to half the distance between SG-E Sea and SG-E Shore, which was the same distance UV Sea and Shore were from the seagrass edge. Dotted lines represent all planned comparisons except between seagrass and unvegetated positions.

27 Sampling was done on eight occasions (four days and four nights) for each patch at each site. Fish samples were collected at each position on each sampling occasion. Fish were sampled using a 1  0.5 m push net with 1 mm mesh, pushed parallel to the patch edge for 5 m. At night, torches were used to determine the seagrass edge and for fish identification, but were turned off during sampling. Sampled fish were anesthetised and preserved in ethanol for later identification and counting.

Seagrass structure and depth Seagrass structure can change across patches and affect fish assemblages (Anderson 2003), potentially confounding the influence of an edge. To document changes in seagrass structure (biomass, length, leaf density, epiphyte biomass) across positions, four haphazardly placed samples of seagrass were collected at the completion of the faunal sampling at each seagrass position within each patch at each site (4 samples  3 positions  4 patches  3 sites = 144 samples). Seagrass samples were taken by cutting all seagrass within a 0.25 m2 quadrat. Water depth was also measured at the point of each fish sample. In the laboratory, macroalgae in the seagrass samples was separated from seagrass. The number and length of leaves was measured for seagrass. Seagrass and algae were weighed after drying to constant weight at 60 C°.

Data analysis Data were assessed for assumptions of homogeneity of variance and normality by viewing box plots and plots of residuals (Quinn & Keough 2002). Where these assumptions were not met, samples were log transformed and reassessed (Quinn & Keough 2002).

Fish Total fish densities, species richness and densities of the most abundant fish species were analysed using four-factor partly-nested analyses of variance (ANOVA) with one between-plot factor and two crossed within-plot factors. Position (7 levels), Site (3 levels) and Time of Day (2 levels) were treated as fixed factors, and seagrass patches (4 in each site) were nested within Sites. Interactions between Patch nested in Site and Time of day, and Patch nested within site and Position have no F-ratio test and were

28 therefore included in the error. UV Shore positions were not taken at patches one and four at Site 1 because these patches were too close to the shore and other suitable patches could not be found. I was not interested in small scale temporal variation; therefore I averaged fish samples across sampling occasions for each patch to simplify the ANOVA model. Specific planned comparisons were run (Figure 2) and Tukey’s tests used to determine differences among sites.

Seagrass structure and depth Variation in seagrass structure, including seagrass biomass, epiphyte biomass, seagrass length and seagrass leaf density across positions was examined using three-factor ANOVAs, Site and Position were fixed factors, while patches were nested within Site.

A four-factor ANOVA was used to determine if depth varied across positions, sites and diel periods. Position, Site and Time of Day were treated as fixed factors, while patches were nested within Site. Specific planned comparisons were used to assess differences between positions (as for fish densities), and Tukey’s tests were used to differentiate among sites.

Results Seagrass structure and depth Seagrass biomass and length varied among positions but with a different pattern at different sites (Table 2). Seagrass was generally longer and more dense and therefore had greater biomass at SG-E Sea and SG Mid at Site 1, at the Mid position at Site 2, and at SG-E Sea at Site 3. SG-E Shore always had the lowest seagrass variables (Figure 3).

There were significant differences in depth among positions that varied with site (F

12,113 = 3.6, p < 0.001), and between times of day that also varied with site (F 2,113 = 10.7, p < 0.001). At all sites and times of the day however, depth was greater at the seaward positions than the middle, which was deeper than the shoreward positions

(Figure 4). Depth at each patch within each site also varied (F 9,113 = 10.5, p < 0.001).

29 Table 2. ANOVA results for seagrass biomass, length, leaf density and epiphyte biomass. Significant results in bold (p < 0.05).

Seagrass Biomass Blade Length Leaf Density Epiphyte Biomass Source df F p F p F p F p Site 2 2.8 0.111 1.8 0.228 9.6 0.006 2.8 0.596 Position 2 27.6 <0.001 35.3 <0.001 11.5 0.001 27.6 0.004 Position x Site 4 4.7 0.009 4.4 0.012 2.2 0.112 4.7 0.276 Patch{Site} 9 6.5 <0.001 13.8 <0.001 2.6 0.010 6.5 <0.001 Position x Patch{Site} 18 2.7 0.001 2.7 0.001 2.8 0.001 2.7 <0.001 Error 108 3 0 300 Biomass ) 2 -

m 200

. g (

s s a

m 100 o i B

200 Length

150 ) m m (

h t 100 g n e L 50

SG-E Sea

SG Mid

) Leaf 2 3 SG-E Shore

m Density

. 0 0 0 ‘ ( 2 y t i s n e D

f 1 a e L

6

) Epiphyte 2

m Biomass

. g (

s 4 s a m o i B

e

t 2 y h p i p E 0 Site 1 Site 2 Site 3 Site

Figure 3. Average seagrass biomass, length, leaf density and epiphyte biomass at each site and position (± 1 SE).

31 100 Site 1

80

60

40

20

100 Site 2

80 UV Sea UV-E Sea 60 ) SG-E Sea m c

( SG Mid

h

t 40 SG-E Shore p

e UV-E Shore D UV Shore 20

100 Site 3

80

60

40

20

0 Day Night Time of Day Figure 4. Average depth at each position within each site (± 1SE).

32 Fish A total of 2 094 fish, from 24 species and 12 families was caught. Larval Atherinidae dominated samples numerically (689 individuals), however, atherinids are highly mobile, transient schooling fish that are not seagrass residents and were therefore removed from further analyses. Stigmatopora nigra was the second most abundant species (377 individuals), followed by S. argus (376), Nesogobius maccullochi (295), Heteroclinus adelaide (125), Sillaginodes punctata (59) and Cristiceps australis (57); these six species represented 92 % of the fish sampled (Table 3) and represent seagrass canopy (S. nigra, S. argus, S. punctata), sand (N. maccullochi) and benthic (H. adelaide, C. australis) species. Seven species were sampled exclusively in seagrass, while only three were sampled only on unvegetated sand, fourteen species were sampled over both habitats, including the six most abundant species (Table 3). Four species (S. nigra, S. argus, N. maccullochi and C. australis) showed a strong response to seagrass edges when analysed separately (Table 6, 7). The vast majority of N. maccullochi (63 %), S. punctata (84 %), Acanthaluteres sp. (92 %) and Gymnapistes marmoratus (90 %) were caught at Site 1. Overall catch per unit effort (CPUE) was low but, due to the sampling technique required to sample fish on a fine spatial scale, and the large portion of samples taken on unvegetated habitats, low CPUE was not unexpected.

33 Table 3. Total abundances of fish sampled at each position at all sites and times.

UV-E SG-E UV-E UV Family Common Name Species UV Sea Sea SG-E Sea SG Mid Shore Shore Shore Total Atherinidae Atherinosoma sp. 0 0 0 0 0 0 1 1 Pikehead hardyhead Kestratherina esox 0 2 2 0 0 1 0 5 Silver fish Leptatherina presbyteroides 0 0 0 0 1 0 0 1 Clinidae Southern crested weedfish Cristiceps australis 0 8 26 11 8 4 0 57 Adelaide weedfish Heteroclinus Adelaide 0 3 52 52 16 2 0 125 Weedfish larvae 1 0 0 0 0 0 0 1 Enoplosidae Old wife Enoplosus armatus 0 0 2 0 0 0 0 2 Gobiidae Bridled goby Arenigobius bifrentatus 0 3 0 0 1 1 0 5 Goby Nesogobius maccullochi 8 33 12 31 127 72 12 295 Monacanthidae Leatherjacket Acanthaluteres sp. 1 4 15 14 13 1 0 48 Six spine leatherjacket Meuschenia freycineti 0 0 3 0 0 0 0 3 Odacidae Little rock whiting Neodax balteatus 0 0 1 0 0 0 0 1 Pleuronectidae Greenback flounder Rhombosolea tapirina 0 0 0 0 5 0 2 7 Scorpaenidae Cobbler Gymnapistes marmoratus 3 5 2 6 9 3 1 29 Scorpidiae Moonlighter Tilodon sexfasciatum 0 0 0 1 0 0 0 1 Sillaginidae King George whiting Sillaginodes punctata 0 8 12 14 22 1 2 59 Spotted pipefish Stigmatopora argus 1 53 193 54 61 13 1 376 Wide body pipefish Stigmatopora nigra 5 108 160 46 33 23 2 377 Halfbanded pipefish Mitotichthys semistriatus 0 0 1 1 0 0 0 2 Pugnose pipefish Pugnaso curtirostris 2 1 2 0 0 0 0 5 Port Phillip pipefish Vanacampus phillipi 0 1 0 1 1 0 0 3 Contusus sp. 0 0 1 0 0 0 0 1 Smooth toadfish glaber 0 1 0 0 0 0 0 1 Total 21 230 484 231 297 121 21 1405 3 4 Fish density Fish density varied for seagrass position and time of day, but both patterns were inconsistent amongst sites (Table 4). At Site 2 and Site 3, density was greater at the UV- E Sea than UV Sea and at Site 1 UV-E Shore was greater than UV Shore (Table 5, Figure 5). When each site was considered individually, fish density was always greater at SG-E Sea than SG Mid but there was no difference between SG-E Shore and Mid (Table 5, Figure 5).

Species richness Species richness differed among positions but not in the same way at each site (Table 4). Within-patch differences were only found at Site 1 (SG Mid > SG-E Shore) and Site 3 (SG-E Sea > SG Mid, Table 5, Figure 5). Although species richness was greater in the seagrass than unvegetated positions and UV-E Sea than the more distant UV Sea at all sites, at Site 2 and Site 3 there was no difference in richness between SG-E Sea and UV-E Sea (Table 5, Figure 5).

Table 4. ANOVA results for fish density and species richness. Significant results in bold (p < 0.05).

Fish Density Species Richness Source df F p F p Site 2 2.1 0.185 3.4 0.078 Position 6 26.5 <0.001 36.1 <0.001 TOD 1 14.3 <0.001 58.0 <0.001 Position*Site 12 3.0 0.001 3.4 <0.001 Position*TOD 6 1.2 0.319 1.0 0.417 Site*TOD 2 5.6 0.005 0.9 0.435 Site*Position*TOD 12 0.8 0.624 0.4 0.965 Patch{Site} 9 3.7 <0.001 3.8 <0.001 Error 113

35 Table 5. Planned comparison results for fish density and species richness at each site pooled across Time of Day. (*= < 0.05, **= < 0.01, ***= < 0.001, > < indicate position with the greatest density).

Fish Density Species Richness Site 1 Site 2 Site 3 Site 1 Site 2 Site 3

Veg Vs Unveg ***> ***> ***> ***> ***> Seagrass

SG-E Sea Vs SG Mid *> **> *> **>

SG-E Shore Vs SG Mid ***<

SG-E Sea Vs SG-E Shore **> ***> Unveg

UV-E Sea Vs UV Sea ***> ***> ***> ***> ***>

UV-E Shore Vs UV Shore *> ***>

Sand Sea Vs Sand Shore **> ***> ***> ***> ***> Mixed

SG-E Sea Vs UV-E Sea ***> *> ***>

SG-E Shore Vs UV-E Shore **> *> ***> ***> *>

Site 1 Site 2 Site 3 5 Fish Density 4 ) 2 - y t 3 i m s

. n 2 UV Sea h e

s UV-E Sea i D f

( 1 SG-E Sea 0 SG Mid s Species Richness s SG-E Shore ) e 2 4 - n UV-E Shore h m

UV Shore

c 3 . i s R e 2 i s c e e i 1 p c s e ( p 0 Day Night Day Night Day Night S Time of Day Figure 5. Fish density (fish m-2) and species richness (species m-2) within each position, at each site during the day and night (± 1 SE).

36 Stigmatopora nigra Stigmatopora nigra densities differed among positions, but patterns varied across sites (Table 6). When each site was considered individually, S. nigra densities were greater at the seaward edge than the middle of seagrass patches at all sites, but there was no difference between the shoreward edge and middle (Table 7, Figure 6). S. nigra used unvegetated edge positions extensively at Site 2, where densities were greater at UV-E Sea than UV Sea, and at Site 3, where densities were greater at UV-E Sea the SG-E Sea (Table 7, Figure 6).

Stigmatopora argus Stigmatopora argus densities also differed among positions, but inconsistently across sites (Table 6). At site 1 and 2, S. argus densities were greater at SG-E Sea than SG Mid and there was no difference between SG-E Shore and SG Mid (Table 7, Figure 6). However, these patterns were not apparent at Site 3, where few S. argus were sampled. S. argus were not sampled in unvegetated positions at Site 1, but at Site 2 densities were greater at UV-E Sea than UV Sea (Table 7, Figure 6).

Table 6. ANOVA results for individual fish species. Significant results in bold (p < 0.05).

Stigmatopora Stigmatopora Nesogobius Cristiceps nigra argus maccullochi australis Source df F p F p F p F p Site 2 0.3 0.718 4.9 0.036 9.6 0.006 4.6 0.042 Position 6 17.4 <0.001 13.7 <0.001 13.1 <0.001 4.7 <0.001 Time of Day (TOD) 1 0.4 0.512 <0.1 0.962 20.1 <0.001 16.8 <0.001 Position*Site 12 4.0 <0.001 3.6 <0.001 1.3 0.231 1.7 0.075 Position*TOD 6 1.7 0.123 0.3 0.912 2.3 0.037 3.3 0.005 Site*TOD 2 1.6 0.203 1.5 0.237 2.9 0.058 1.3 0.272 Site*Position*TOD 12 1.5 0.130 0.7 0.706 1.0 0.483 1.4 0.164 Patch{Site} 9 3.6 0.001 2.9 0.004 1.6 0.114 1.1 0.360 Error 113

37 Nesogobius maccullochi Densities of the goby, Nesogobius maccullochi, varied with site, and differed among positions but in a different way during the day and night (Table 6). During both the day and night, more N. maccullochi were sampled at the shoreward seagrass edge than the middle. At night, N. maccullochi were always sampled in greater densities at the unvegetated edge positions than the more distant unvegetated positions (Table 7, Figure 6). During the day more N. maccullochi were sampled at SG-E Shore than UV-E Shore, but there was no difference during the night.

Site 1 Site 2 Site 3 1.2 Stigmatopora nigra

0.8

0.4

1.8 Stigmatopora argus

1.2 ) 2 - m

. 0.6 UV Sea h

s UV-E Sea i f (

SG-E Sea y t i 2.0 SG Mid s Nesogobius maccullochi n SG-E Shore

e 1.6

D UV-E Shore 1.2 UV Shore 0.8 0.4

0.6 Cristiceps australis

0.4

0.2

0 Day Night Day Night Day Night Time of Day

Figure 6. Average density at each site of Stigmatopora nigra and S. argus at each position pooled over diel periods (no significant interaction with time of day), and Nesogobius maccullochi and Cristiceps australis at each position during each diel period (±1 SE).

38 Cristiceps australis The weedfish, Cristiceps australis, was sampled in greater densities at SG-E Sea than both SG Mid and SG-E Shore during the night but not during the day (Table 6, 7, Figure 6).

39 Table 7. Planned comparison results for individual fish species within each site for Stigmatopora nigra and S. argus (pooled across time of day) and time of day for Nesogobius maccullochi and Cristiceps australis (pooled across site). ( *= < 0.05, ***= < 0.01, ***= < 0.001, > < indicate position with the greatest density).

Stigmatopora Stigmatopora Nesogobius Cristiceps nigra argus maccullochi australis Site 1 Site 2 Site 3 Site 1 Site 2 Site 3 Day Night Day Night

Veg Vs Unveg ***> ***> ***> ***> ***> ***> ***> ***> ***> ***> Seagrass

SG-E Sea Vs SG Mid **> *> ***> ***> **> ***>

SG-E Shore Vs SG Mid ***> ***>

SG-E Sea Vs SG-E Shore ***> **> **> **> ***> ***< ***< ***> Unveg

UV-E Sea Vs UV Sea **> ***> **> *>

UV-E Shore Vs UV Shore **>

UV-E Sea Vs UV-E Shore *> ***> *> *< Mixed

SG-E Sea Vs UV-E Sea ***> *> ***> **> *<

SG-E Shore Vs UV-E Shore **> ***> 4 0 Relationships between fish densities and seagrass structure Fish density, both overall and of individual species, showed different patterns to seagrass structure. For example, at Site 1 there was no difference in seagrass biomass at SG-E Sea and SG Mid but there were clearly fewer fish at the middle (Figure 3, 5). Similarly, at Site 2 seagrass biomass was greatest at the middle but Stigmatopora nigra densities were much greater at the seaward edge (Figure 3, 6).

Discussion Fish assemblages can vary according to within-patch location, but the presence of an edge effect in seagrass habitats has been inconclusive. While Uhrin & Holmquist (2003) found fish densities to be greatest 10 m from the seagrass edge, and others found changes in individual species densities at the seagrass edge (Hovel et al. 2002, Jelbart et al. 2006), most studies have found no difference in fish density between the edge and interior (Connolly & Hindell 2006). I found both overall and individual fish densities were greater at seagrass edges than the middle. Fish density was greater at the seaward edge than the middle, but not between the shoreward edge and middle. Similarly, the pipefish Stigmatopora nigra and S. argus, and the weedfish Cristiceps australis were much more common at the seaward edge than the seagrass middle, although C. australis only showed this pattern during the night. The goby Nesogobius maccullochi, by contrast, was more abundant at the edges in the shoreward direction than at the seagrass middle. Patterns of both overall and individual within-patch fish distributions provide support for edge effects, contrasting with previous work that has found little evidence of seagrass edge effects (Connolly & Hindell 2006).

Seagrass edges may provide pipefish with an advantage over other microhabitats within a seagrass patch. Pipefish feed on planktonic copepods and other small (Kendrick & Hyndes 2005), which may be more plentiful at the seagrass edge (Tanner 2005), suggesting that pipefish may use the edge because there is greater food availability.

Unvegetated habitats support fewer fishes than seagrass habitats (Ferrell & Bell 1991), but use of sand at the seagrass/sand interface by seagrass associated fish is not well

41 known. Anderson (2003) found that sand-associated fishes such as flatheads and stingarees were more common in close proximity to structure than completely unvegetated habitats, supporting the findings of Ferrell & Bell (1991) that non-seagrass fishes are more abundant in sand within 10 m of seagrass than either seagrass, or sand greater than 100 m away. I found that sand habitats immediately adjacent to seagrass patches were used extensively by fish, particularly in relation to more distant unvegetated habitats. Fish density and number of species were greater over sand at the edge of seagrass than the more distant unvegetated sand, and the six most abundant species were all sampled at the sand edge. Additionally, seagrass-associated species sampled in this study, while showing no difference in distribution within seagrass patches, were sampled in greater densities in adjacent sand than distant sand (Smith unpubl. data). At Site 1, however, few fish were sampled at the sand edge. Fish prey items (meiofaunal crustaceans) are lower in abundance (Jenkins & Hamer 2001) and predation pressure greater (Hindell et al. 2002) in unvegetated habitats at Site 1 than other seagrass associated sites within Port Phillip Bay, suggesting predation or food availability are reducing fish utilisation of the sand edges at this site. Sand at the seagrass edge may provide a place for fish to forage while still within close proximity to the safety of seagrass, that more distant sand habitats cannot provide (Orth et al. 1984, Anderson 2003), however, more experimental work is needed to determine the importance of unvegetated habitats directly next to seagrass patches.

Fish assemblages can change with diel cycles (Gibson et al. 1996, Nagelkerken et al. 2000), but within-patch diel variations have only recently been investigated. Jackson et al. (2006b) found that the influence of seagrass landscape attributes, including edge effects, on fish assemblages change with diel cycles. In the current study, although more fish were sampled at night, time of day had little effect on fish microhabitat choice for most species. The one exception was Cristiceps australis, which was more abundant at the seaward seagrass edge than other positions at night but not during the day. Movement to the seagrass edge at night may reflect changes in predator abundance or food availability, but too little is known of the ecology of C. australis to draw firm conclusions for the species.

42 Changes in seagrass structure within seagrass patches can influence fish assemblages (Bologna & Heck 2002), potentially explaining edge effects within seagrass habitats. The response of fish in relation to seagrass structure (biomass, length, density, epiphytes) has been well studied (Orth et al. 1984, Anderson 2003, Hyndes et al. 2003, Jackson et al. 2006b), increasing or decreasing fish density depending on species. I found seagrass structure varied across and within patches, possibly explaining differences in fish assemblages within patches. Differences in structure between the seaward edge and middle varied inconsistently across sites. Fish densities between the seaward edge and middle, however, were consistent across sites, suggesting that within- patch differences in structure are not determining fish assemblages. Although depth can influence the distribution of a variety of fishes (Jackson et al. 2006b), the depth at the seaward edge and middle was always much greater than the seagrass canopy and unlikely to influence small seagrass fish sampled in this study. The shoreward seagrass edge always had lower seagrass density, biomass and length, and was shallower than the middle of a patch. This might be unfavourable for some species (Jackson et al. 2006b) and potentially mask any edge effect at the shoreward edge. Conversely, Nesogobius maccullochi showed a preference for the shoreward edge, which may reflect a preference for sparse, short seagrass and shallow water, and not an edge effect. Ries & Sisk (2004) emphasised that variations in patch quality will cause variability in species responses to edges. I found that fish showed a strong response to the seaward seagrass edge but little at the shoreward edge. Differences in seagrass structure (density, length, biomass) and depth are the probable cause of the contrasting responses at the seaward and shoreward edges, but could not explain the difference between the middle and seaward edge.

Edges are a common feature of landscapes that influence environmental factors (Murcia 1995), species interactions (Fagan et al. 1999) and trophic transfer (Cadenasso et al. 2003). Seagrass habitats produce a distinct boundary with unvegetated sand habitats, producing changes in water flow (Bologna & Heck 2002), sediment particle size (Fonseca & Fisher 1986) and interactions between seagrass and sand predators and prey (Bologna & Heck 1999). Seagrass edges are thought to represent a reflective or absorptive edge for seagrass specialists because of the lack of shelter outside seagrass

43 patches. At both Site 2 and 3, however, I found strong use of both the seaward seagrass and sand edges by fish, which may reflect a spill over effect, where aggregated individuals may ‘spill’ from a preferred habitat onto a non-preferred habitat (Ries & Sisk 2004). Alternatively, the sand edge may provide a complementary habitat (Ries & Sisk 2004) where water currents, food availability, foraging success and predation rates are beneficial.

Previous work on seagrass edge effects have produced inconsistent results (Connolly & Hindell 2006). I found strong evidence of an edge effect in seagrass patches that could be both explained (shoreward edge) and not explained (seaward edge) by seagrass structure and depth. Fish used sand adjacent to seagrass extensively, further supporting the importance of seagrass edges to seagrass fish but mechanisms establishing edge effects have only been hypothesised and need to be tested. As seagrass habitats are becoming increasing fragmented around the world, there is an increased need to establish the effects and implications of edges to develop and improve management of seagrass habitats.

44 3. Seagrass patch size affects fish responses to edges

Published: Smith TM, Hindell JS, Jenkins GP, Connolly RM (2010) J Anim Ecol, 79: 275-281

ABSTRACT 1. Patch area and proximity of patch edge can influence ecological processes across patchy landscapes and may interact with each other. Different patch sizes have different amounts of core habitat, potentially affecting abundances at the edge and middle of patches. In this study we tested if edge effects varied with patch size. 2. Fish were sampled in ten various-sized seagrass patches (114 – 5 934 m2) using a small (0.5 m2) push net in three positions within each patch: the seagrass edge, 2 m into a patch, and in the middle of a patch. 3. The two most common species showed an interaction between patch size and the edge – interior difference in abundance. In the smallest patches, pipefish (Stigmatopora nigra) were at similar densities at the edge and interior, but with increasing patch size, the density at the edge habitat increased. For gobies (Nesogobius maccullochi), the pattern was opposite, having similar densities at the edge and middle in small patches, but greater density in the middle as patch size increased. 4. This is the first example from a marine system of how patch size can influence the magnitude and pattern of edge effects. 5. Both patch area and edge effects need to be considered in the development of conservation and management strategies for seagrass habitats.

Keywords: edge effects, patch size, landscape structure, seagrass, fish,

45 Introduction

Habitat edges and patch size are key elements in landscape ecology, directly affecting animal abundances and distributions (Murcia 1995, Bender et al. 1998, Fletcher et al. 2007). Edges can influence physical parameters and species interactions, increasing and decreasing animal densities (Paton 1994, Fagan et al. 1999, Ries et al. 2004). The role edges play in determining faunal distributions vary with temporal scales, patch quality, mechanisms of patch formation and contrast between patches and the surrounding matrix (Ries et al. 2004, Ries & Sisk 2004, Ewers & Didham 2006). Patch area is included in many landscape studies, but it is linked to habitat edges, and the two are often confounded and difficult to isolate (Bender et al. 1998, Parker et al. 2005, Fletcher et al. 2007). The few studies that have investigated the interaction between edge effects and patch size have generally found an interaction between them (Didham et al. 1998, Barbosa & Marquet 2002, Ewers et al. 2007), suggesting that the strength and direction of edge effects may be related to patch size (Figure 7). The relationship between patch size and edge is important to ecologists because small patches generally have a greater edge to area ratio and less core habitat than larger patches, and the distance to edges is less in small patches, suggesting that edge effects could drive area effects (Malcolm 1994, Fletcher 2005, Fletcher et al. 2007). The relationship is fundamentally important in the conservation, restoration and management of habitats suffering fragmentation through anthropogenic activities (Ries et al. 2004, Ewers et al. 2007).

Seagrass is a common nearshore ecosystem with a propensity to form patches of various sizes interspersed with unvegetated sand (Robbins & Bell 1994) and that is susceptible to anthropogenic disturbance (Duarte 2002, Orth et al. 2006a). The influence of seagrass edges and patch area on fish densities and distributions in seagrass is unclear, with the majority of species showing no apparent effect (Bell et al. 2001, Connolly & Hindell 2006). Recent work, however, suggests that edges play a role in structuring the composition of seagrass fish assemblages (Jelbart et al. 2006, Smith et al. 2008), and indicates that patch size may explain some of the inconsistencies among studies (Jelbart et al. 2006). Jelbart et al. (2006) found that fish species richness did not differ between the interior and edge of small patches, but was lower at the edge of large patches; suggesting that small patches may have no suitable core habitat for fish, and that edge

46 effects may drive area effects in seagrass patches. However, patch size and edge effects did not interact to influence fish density; only one species showed an edge effect, while eight showed an area effect. This indicates that edge and area effects may function independently of each other to influence the abundance and distribution of individual fish species. Inconsistencies among the results of studies that have assessed edge effects in seagrass may be related to the size of the patches sampled. For example, Smith et al. (2008) found that total fish density and the density of three individual species showed an edge effect, but they sampled smaller patches than Jelbart et al. (2006).

Edge Effect Strength (Edge – Interior) +ve Edge Effect y t i s n e D

h s i Interior

F 0 Effect

Interior 0

y Effect t i s n

e Edge D

h s i -ve F Effect Patch Area Patch Area

Figure 7. Schematic representation of changes in strength and direction of edge effects as patch size increases if edge effects and patch area interact. Diagrams on the left show density differences at the edge and middle as patch area increases, and the consequent changes in edge effect strength and direction on the right.

The strength of edge and patch effects on faunal composition and abundances can be affected by the surrounding landscape (Ewers & Didham 2006, Tanner 2006). Landscape variables such as distance to nearest neighbour, to shore, and to other patch

47 types, can affect species abundance patterns (Heithaus et al. 2006, Jelbart et al. 2007), which may in turn affect abundance at the patch edge. In addition, changes to the attributes of seagrass patches may influence fish abundance within and between patches (Boström et al. 2006). Indeed, variation in the structure and depth of seagrass can affect the distribution of fish within a patch (Jackson et al. 2006a, Smith et al. 2008). Differences in within patch seagrass structure therefore also need to be considered when assessing edge and area effects.

The definition of seagrass patch edge/interior varies among studies, often varying with study organism and patch size. The seagrass interior can be represented as: a fixed distance into a patch (e.g. 5 and 10 m Uhrin & Holmquist, 2003; > 10 m Bologna 2006, 6 - 8 m Jelbart et al. 2006), which is consistent regardless of patch size; or at the halfway point across each individual patch (e.g. Bowden et al. 2001, Smith et al. 2008). Variation in definitions of the seagrass interior may explain some of the inconsistencies among studies investigating seagrass edge effects.

I hypothesised that, in small patches, edge effects are small because edge and core habitats are similar, and that as patch size increases, edge effects will increase as the edge and core become more dissimilar. I tested this by determining if differences in fish densities between the edge and interior of seagrass patches are influenced by patch size. I also determined if fixed or proportional distances into patches differ in representing the interior of patches, and tested whether seagrass structure explained fish variables not responding to patch size.

Methods

Study Site Sampling was done in May and October 2006 at Blairgowrie on the south-eastern coast of Port Phillip Bay, a large (2,000 km2) embayment in Victoria, Australia. Blairgowrie supports patches of the seagrass Heterozostera nigricaulis of various shapes and sizes, interspersed by 10s of metres of sand, to a depth of 1.5 metres. In recent years, the area of seagrass has declined (Ball et al. 2006), and seagrass habitat has become increasingly patchy.

48 Table 8. Landscape characteristics of each patch sampled.

Area Nearest Distance to Average Depth Patch (m2) Neighbour (m) Shore (m) (cm ± SE) 1 5934 4 30 53 (4) 2 3098 43 19 52 (4) 3 1126 4 91 49 (3) 4 522 17 94 74 (4) 5 485 17 108 55 (4) 6 461 20 21 34 (3) 7 413 29 99 57 (4) 8 269 10 46 60 (4) 9 239 30 56 50 (4) 10 114 8 66 45 (3)

Sampling Methods Ten patches were sampled within an hour of low tide (Table 8) during two seasons: autumn (May) and spring (October). Samples were taken at three positions within each patch; the seagrass edge, within one metre of the seagrass/sand interface (edge); fixed seagrass interior, two metres from the seagrass edge (2 m); and, half way across the seagrass patch at the proportional seagrass interior (middle), between 2 and 60 m from the seagrass edge. During each season (autumn, spring), samples were collected on eight occasions (four day, four night). To limit disturbance of nearby positions, complete sampling of each patch was done over two sampling occasions (i.e. 2 days and nights each season), where only one or two randomly chosen positions in each patch were sampled on each occasion. At the smallest patch, the middle and 2 m positions were located at the same distance into the patch.

Fish were caught using a 1  0.5 m push net, with 1 mm mesh, which was pushed parallel to the patch edge for a distance of 5 m. All fish were identified and counted. Seagrass structure (above-ground biomass, shoot length, density, epiphyte biomass) was quantified by taking four seagrass samples at each position within each patch during each season (4 samples  3 positions  10 patches  2 seasons = 240 samples). Seagrass samples were taken by cutting all seagrass within a 0.25 m2 quadrat. The number of

49 seagrass leaves in each sample was counted to estimate seagrass leaf density, and the length of ten randomly selected leaves measured. Seagrass and epiphytic algae biomass were measured by weighing these plant components separately after drying to constant weight at 60 °C. Epiphytic algae consisted of small filamentous and thin sheet-like red and green algae which contributed little to the structure of the meadows.

Water depth was measured at the point of each fish sample. Seagrass patch area, distance to shore and distance to nearest neighbour for each patch was measured on rectified aerial photographs using GIS software. Distance to shore was measured as the distance from the high tide mark to the centre of the patch, and nearest neighbour as the smallest distance between seagrass patches.

Data analysis Any interaction between sampling positions and patches would suggest that variation between patches may be affecting edge effects. To assess if positions and patches interacted, species richness and the density of the two most abundant species were each analysed using a three-factor, repeated measures ANOVA. Separate analyses of the same test were done to compare the positions: 1) edge and 2 m, and 2) edge and middle. Season, Time of Day and Patch were treated as between patch subjects, where Season (autumn, spring) and Time of Day (day, night) were treated as fixed factors and Patch as a random factor. Position was treated as the within subjects factor. Data were assessed for assumptions of homogeneity and normality by viewing box plots and plots of residuals; where these assumptions were not met, data were log transformed and reassessed (Quinn & Keough 2002).

Where there was a significant interaction between position and patch, the difference in the response variable between positions (i.e. edge - 2 m or edge - middle) was tested against predictor variables (i.e. patch area and other seagrass measures); depth and seagrass density were analysed by first calculating differences between positions (i.e. edge - 2 m or edge - middle). Stepwise multiple linear regressions were used to assess the relative contribution of patch characteristics in determining variation in abundances of fish across positions and patches. Where there were significant differences among

50 patches (and no interaction), regression analyses were done between predictor variables and fish densities for each patch. Patch area, distance to shore, distance to nearest neighbour, depth, seagrass biomass, density, shoot length and epiphyte biomass were used as predictor variables in regression analysis. Collinearity was tested by doing correlations of predictor variables. Where r values were greater than 0.3, the predictor variable that was deemed to have the least relevance to the primary focus of the study was removed from multiple regression analyses. There is no convention in the statistical literature determining correlation values that contribute to collinearity, and therefore r values greater than 0.3 were chosen as a conservative measure. Distance to shore, seagrass biomass, shoot length and epiphyte biomass were correlated with other predictor variables (r values greater then 0.3) and were removed from analyses. For the same reason, nearest neighbour data were removed from all analyses except for spring densities, and depth was removed from analyses of interactions between position and patch. Therefore, only patch area, seagrass density and, nearest neighbour in spring, and depth when there was no interaction between patch and position, were used as predictor variables in regression analyses.

Results

Fifteen species of fish from nine families were sampled. During autumn and spring, 802 fish from 9 species, and 961 fish from 14 species, respectively, were sampled. The pipefish Stigmatopora nigra (Kaup) was the most abundant species (904 individuals), followed by the goby Nesogobius maccullochi (Hoese & Larson)(425). These 2 species contributed 70 % of all fish sampled, and represent species with affinities for seagrass canopy (S. nigra) and unvegetated sand (N. maccullochi).

Species richness Species richness edge effects did not vary with patch size, but increased with seagrass density. Species richness varied inconsistently amongst patches between seasons at the edge and middle positions (Table 9), but predictor variables explained little of this variation. At the edge and 2 m positions, there was a highly significant difference in species richness amongst patches (Table 10) and seagrass density was positively correlated with species richness (p = 0.033), explaining 38 % of the variability across

51 patches (Figure 8). Species richness did not vary between the edge and middle positions (Table 9) or between edge and 2 m positions (Table 10). A greater number of species was sampled at the edge and middle positions at night and in spring than autumn (Table 9, Figure 10), and at the edge and 2 m positions at night in spring (Table 10, Figure 10).

Table 9. Repeated measures analysis between the edge and middle for species richness, Stigmatopora nigra and Nesogobius maccullochi. Significant results in bold (p < 0.05). Species Stigmatopora Nesogobius Between Subjects Richness nigra maccullochi Source df F p F p F p Season (S) 1 28.1 <0.001 7.0 0.026 10.2 0.011 Time of Day (TOD) 1 13.4 0.005 0.2 0.664 4.7 0.059 Patch (P) 9 11.5 <0.001 18.3 <0.001 6.2 <0.001 S x TOD 1 3.7 0.085 2.5 0.145 0.9 0.355 S x P 9 2.5 0.021 3.4 0.004 1.2 0.340 TOD x P 9 1.6 0.161 0.9 0.521 2.8 0.012 S x TOD x P 9 1.1 0.386 0.6 0.748 0.6 0.772 Error 40

Within Subjects Source df F P F p F p Position 1 0.1 0.761 10.6 <0.001 0.9 0.380 Position x S 1 4.2 0.071 2.0 0.138 0.9 0.364 Position x TOD 1 0.4 0.522 1.6 0.212 0.1 0.716 Position x P 9 0.9 0.563 3.0 <0.001 2.2 0.042 Position x S x TOD 1 <0.1 0.847 4.8 0.009 0.4 0.529 Position x S x P 9 0.6 0.779 1.3 0.176 0.9 0.510 Position x TOD x P 9 0.8 0.648 0.8 0.789 0.5 0.852 Position x S x TOD x P 9 0.5 0.835 0.9 0.544 1.1 0.404 Error 40

52 Table 10. Repeated measures analysis between the edge and 2 m for species richness, Stigmatopora nigra and Nesogobius maccullochi. Significant results in bold (p < 0.05). Species Stigmatopora Nesogobius Between Subjects Richness nigra maccullochi Source df F p F p F p Season (S) 1 62.5 <0.001 5.5 0.044 5.8 0.040 TOD 1 16.8 0.003 0.0 1.000 8.6 0.017 Patch (P) 9 11.5 <0.001 37.2 <0.001 2.4 0.030 S x TOD 1 8.5 0.017 3.7 0.088 0.5 0.491 S x P 9 1.3 0.260 10.5 <0.001 1.2 0.316 TOD x P 9 1.1 0.371 0.8 0.603 0.9 0.560 S x TOD x P 9 0.4 0.934 1.3 0.258 0.9 0.517 Error 40

Within Subjects Source df F p F p F p Position 1 <0.1 0.842 0.5 0.497 0.0 1.000 Position x S 1 0.3 0.572 0.1 0.747 0.8 0.382 Position x TOD 1 1.0 0.356 2.4 0.159 0.0 1.000 Position x P 9 1.6 0.147 1.6 0.162 2.1 0.049 Position x S x TOD 1 <0.1 0.924 8.7 0.016 0.7 0.423 Position x S x P 9 1.1 0.405 0.6 0.789 0.9 0.562 Position x TOD x P 9 0.6 0.754 0.4 0.939 0.9 0.562 Position x S x TOD x S 9 0.7 0.668 0.3 0.980 1.1 0.363 Error 40

53 )

2 4 - m

. s e i c

e 3 p s (

s s e n

h 2 c i r

s e i c e

p 1 S 6000 9000 12000 15000 Seagrass density (leaves. m-2)

Figure 8. Relationship between mean species richness and seagrass density across patches at the seagrass edge and 2 m (pooled).

Pipefish Densities of Stigmatopora nigra at patch edges relative to the middle of patches increased as patch size increased. Differences in S. nigra densities between the edge and middle positions varied among patches (Position x Patch interaction, Table 9); 56% of this variation was explained by patch size (p = 0.041), with greater densities at the edge of large patches (Figure 9). Stigmatopora nigra densities were greater at edge than middle positions during the day in autumn (Table 9, Figure 10), and at 2 m than edge positions at night in autumn (Table 10, Figure 10), but there was little difference for other combinations of season and time of day. Densities of S. nigra at the middle and 2 m positions also varied amongst patches, but this variation depended on season (Table 10). No predictors significantly contributed to patch differences during autumn. In spring, however, depth explained 47 % of S. nigra variation among patches (p = 0.029).

Gobies Densities of Nesogobius maccullochi at the patch edges decreased relative to the middle of patches as patch size increased. Differences in densities of N. maccullochi between edge and middle positions varied among patches (Table 9); 42 % of this variation was explained by patch area (p = 0.026), with greater densities in the middle of large patches

54 (Figure 9). Densities of N. maccullochi at both the edge and middle positions, and edge and 2 m were greater in spring than autumn (Table 9, 10). At the edge and 2 m positions, densities of N. maccullochi varied during diel periods and among patches (Table 10), differences among patches were not explained by patch area, water depth or seagrass density. 3 (a) ) 2 - 2 m

. h s i f ( 1 y t i s n e

D 0

1 (b) ) 2 - m

. 0 h s i f (

y t i s n

e -1 D

2.0 2.5 3.0 3.5 4.0 2 Log10 patch area (m )

Figure 9. Relationship between the difference in density at the edge and interior (edge – interior), and seagrass patch area (log10) for (a) the pipefish Stigmatopora nigra and (b) the goby Nesogobius maccullochi. Data pooled over Season and Time of Day.

55 4 Stigmatopora nigra Edge 2 m 3 Middle 2 - m . h s

i 2 F

1

0 Day Night Day Night Autumn Spring

Figure 10. Average Stigmatopora nigra density at the edge, 2 m and middle during each season and time of day (±1 SE).

Discussion

Edge effects and patch size are key elements of landscape ecology, but they potentially interact and their influences can be difficult to separate (Fletcher 2005, Parker et al. 2005, Fletcher et al. 2007). I have shown that the strength and direction of edge effects on two common fish species in seagrass patches could be explained by patch size, which has a positive (for Stigmatopora nigra) and negative (for Nesogobius maccullochi) effect on densities at the edge. My findings contrast with those of previous work on seagrass fish, which found that species richness interacted with patch size, but that the densities of S. nigra (Jelbart et al. 2006) and invertebrates (Tanner 2006) did not. Patch size and edge width were smaller in this study than in Jelbart et al. (2006), and possible consequences are discussed below. My findings are, however, consistent with terrestrial studies that have found that edge effects and patch size interact to influence abundance (Didham et al. 1998, Barbosa & Marquet 2002, Ewers et al. 2007). Changes in the strength of edge effects as patch size increases suggest that edge effects are driving area effects (Parker et al. 2005, Ewers et al. 2007, Fletcher et al. 2007). The lack of an edge effect for both S. nigra and N. maccullochi in small patches supports the

56 hypothesis that, in small patches, there is little core habitat to sustain differences in densities between the edge and interior (Ewers et al. 2007). In contrast, in larger patches where there is more core habitat, and a greater distance between the edge and middle positions to promulgate changed conditions, a larger edge effect can result (Barbosa & Marquet 2002, Ewers et al. 2007). The detection of few inconsistent differences in species abundance between positions at the edge and 2 m into seagrass patches lends further support to the theory that small patches provide little or no core habitat for fish.

Edge effects are known to vary in marine and terrestrial systems (Murcia 1995, Ries et al. 2004, Ewers & Didham 2006), including seagrass habitats (Boström et al. 2006, Connolly & Hindell 2006). Differences in patch conditions and attributes can explain some of the variability amongst studies (Donovan et al. 1997, Ewers & Didham 2006), and the interaction between edge and patch area is an important component of this variability (Jelbart et al. 2006). Differences in densities of Stigmatopora nigra between the middle and edges of seagrass patches were greater in large than small patches (i.e. there were relatively more fish at the edge in large patches than in small patches). This contrasts with Jelbart et al. (2006), who found no difference in S. nigra densities between the edge and middle of seagrass patches regardless of patch size. Small patches in their study were similar in size to my large patches; it would therefore be expected that greater densities of S. nigra would be sampled at patch edges in their study. Inconsistencies between patterns of S. nigra distributions in Jelbart et al. (2006) and my study may relate to difference in patch characteristics (i.e. seagrass species, length, water depth), landscape variables (i.e. patch location, nearest neighbour) or sampling position within patches. Jelbart et al. (2006) considered the edge to be 4 m wide, and thus may have included some interior habitat in edge samples. In contrast to S. nigra, densities of the goby Nesogobius maccullochi increased in the patch middle relative to the edge as patch size increased. Such a pattern is unusual because N. maccullochi is generally associated with shallow sand habitats (Hoese & Larson 2006) and Smith et al. (2008) found N. maccullochi densities greater at the seagrass edge than middle. Although seagrass density differences could not explain N. maccullochi patterns in large patches, there may have been areas of unvegetated sand or sparse seagrass in large patches that were not considered in seagrass samples, potentially explaining greater N.

57 maccullochi densities at the interior of large patches. Large-scale landscape characteristics such as position within a landscape, the role of corridors and patch aggregation have only recently been considered in seagrass ecosystems (Boström et al., 2006), and may provide insights into inconsistencies among edge studies.

Studies investigating edge effects in seagrass have used both fixed (Uhrin & Holmquist 2003, Bologna 2006, Jelbart et al. 2006) and proportional (Bowden et al. 2001, Smith et al. 2008) distances into seagrass to represent the interior of patches. Abiotic and biotic differences may occur far into habitat interiors (Barbosa & Marquet 2002), and the influence of these factors may be overlooked if fixed positions are not placed far enough into patch interiors. I found little difference in fish density between the fixed distance interior (i.e. 2 m into a patch) and the patch edge. The lack of difference between the edge and fixed interior possibly points to a deficiency in the approach of choosing a fixed distance to represent the interior of a patch; although at only 2 m into the patch, the fixed position adopted in this study may still be representative of edge habitat. However, choosing the alternative approach and representing the middle of patches using proportional distances into patches, provides no controls for the influence of patch area on edge effects (i.e. no difference in small patches with little core habitat). My results suggest a constant increase in edge effects as patch size increases, although I may not have sampled enough patches, or at enough positions within patches, to detect any exponential patterns in edge effects. To separate the effects of area and edge, (Fletcher et al. 2007) suggest sampling at multiple distances into patches of distinct categorical sizes (e.g. small, medium, large). Such sampling regimes, however, may prove difficult in sampling mobile species in small seagrass patches, which are often too small to sample at more than one or two distances into a patch. Novel, fine-scale sampling techniques and regimes (e.g. Minello & Rozas, 2002) may be required to further elucidate edge and area effects in seagrass.

Distribution of resources across habitat patches affects the magnitude of edge effects (Ries et al. 2004). Food availability has been suggested as an important factor in explaining the distribution and abundance of Stigmatopora nigra within seagrass patches (Smith et al. 2008, Macreadie et al. 2009). Stigmatopora nigra are a permanent

58 seagrass species that ambush small benthic and planktonic copepod crustaceans (Steffe et al. 1989, Kendrick & Hyndes 2005) that vary in abundance within seagrass beds (Warry et al. 2009). The reduced water movement at the interior of seagrass patches (Peterson et al. 2004), potentially limits the flow of planktonic copepods from outside to the interior of the patch, and, therefore, may reduce the abundance of prey for S. nigra; reduced prey densities may explain, or contribute to, differences in S. nigra abundance between the edge and interior of large patches. In contrast, small patches may not support enough habitat and structure to restrict flow and subsequent food supply into patch interiors. Similarly, Nesogobius maccullochi densities were greater in the middle of large patches than at the edge, but showed little difference in small patches. Gobies are a common prey item of piscivorous predators in Port Phillip Bay (Hindell et al. 2000), and may find greater refuge in the middle of large patches, but in small patches there is little difference between the edge and middle. The less significant differences in smaller patches may thus reduce the magnitude of edge effects on (Barbosa & Marquet 2002).

The relationship between edge and area becomes increasingly complicated in small patches, and a given distance (whether fixed or proportional) into a small patch may not behave in the same manner as in a larger patch (Ewers et al. 2007). Locations within small patches are closer to more edges than in large patches, and are therefore exposed to the effects of multiple edges (Malcolm 1994, Fletcher 2005). The role of multiple edges in small patches may be additive, in which case edge effects may increase exponentially (Malcolm 1994, Ewers et al. 2007). Small patches can also encourage species aggregation (Cronin 2009), further complicating the relationship between edge and area. Patch shape can also affect animal distributions and abundance (Jelbart et al. 2006), and is exaggerated in small patches. Although not considered in our study, patch shape effects may have contributed to differences in edge effects between patches.

Edge effects can be driven by differential predation rates between edge and interior habitats (Lecomte et al. 2008), and the abundances of predators and their prey can change over time (Young & Mitchell 1994, Meyer et al. 2001, Ries et al. 2004). In seagrass habitats, a small number of species have shown different responses to edges

59 over time (e.g. Hovel et al. 2002, Smith et al. 2008). I found that season and time of day had little effect on the within-patch variation of seagrass fishes. Relative differences in species richness and goby density between positions in seagrass patches did not change over time, while the density of Stigmatopora nigra varied inconsistently between positions over time, but maintained consistent patterns across patches. Stigmatopora nigra predator and prey abundances have been shown to vary over time (Jenkins & Hamer 2001, Hindell 2006), and may contribute to the temporal variance of edge effects on S. nigra.

Patch size can determine the presence and strength of edge effects. Interactions between edge effects and patch area can influence the interpretation of landscape studies, highlighting the need for carefully planned experiments (Fletcher et al. 2007). Habitat fragmentation reduces patch size and increases the amount of edge habitat, making the relationship between edge and area important for conservation, restoration and management, particularly for species favouring interior habitats that are likely to be most affected. Indeed, understanding at what size a patch becomes entirely edge would be a powerful management tool and predicting the smallest viable patch size for a species would allow managers to set trigger points preventing lose of essential core habitat. For instance, patch sizes smaller than 400 m2 may reduce the population of gobies at Blairgowrie. Further investigations to establish when an edge habitat becomes an interior habitat and the underlying mechanisms affecting the relationship between edge and area will be important for continuing conservation and management of fragmented ecosystems.

60 4. Fine-scale spatial and temporal variations in diets of the pipefish Stigmatopora nigra within seagrass patches in Port Phillip Bay, Victoria, Australia

ABSTRACT: Edge effects are a prominent feature in ecology and can be attributed to changes in food abundances. Seagrass beds are a common nearshore habitat harbouring rich and diverse assemblages of fish. Pipefish from the Stigmatopora are strongly associated with, and very common in seagrass habitats. Previous work has shown the pipefish Stigmatopora nigra are most abundant at the edge of seagrass beds, where copepods, their predominant prey, are found in highest abundances. In an earlier study (Smith et al. 2010), abundances of S. nigra collected from the edge and middle of seagrass patches were greater at the seagrass edge during the day in autumn, but not at any other time. I hypothesised that if prey abundances in S. nigra diets were greater at the seagrass edge than middle during the day in autumn, but not at any other time, prey availability, inferred from diets, was causing differences in S. nigra abundances at the edge and middle. Gut analyses were done to determine if there was any difference in diet between the edge and middle of patches, and if patterns of prey consumption varied with seasons, time of day or seagrass structure. There was no evidence to support the hypothesis and there was little difference in prey consumed by S. nigra at the edge and middle of patches, regardless of time of day or season. Seagrass density however influenced S. nigra diets. Harpacticoid copepods were more abundant, and, planktonic copepods less abundant, in S. nigra guts as seagrass density increased. Stigmatopora nigra diets may change with seagrass structure relative to changes in prey availability. The lack of any difference in prey eaten by S. nigra at the edge and middle of seagrass patches suggests that food is unlikely to be determining S. nigra distribution patterns at the edge and middle of patches. Alternatively, any edge effects in diet are being masked by other factors such as seagrass structure.

Keywords: Stigmatopora nigra, gut analysis, seagrass, edge effects

61 Introduction Species distributions are governed by a variety of factors including resource allocation, species interactions and abiotic conditions (Begon et al. 1996). At habitat edges, changes in physical structure, and subsequent changes in environmental variables can alter the distribution of resources (Murcia 1995, Ries & Sisk 2004). Changes in resource distribution at habitat edges can have direct (preferred habitat) and indirect (change in interactions) effects on fauna (Murcia 1995). Edge effects and their impact on species distribution is a fundamental concept in landscape ecology, and with increasing habitat fragmentation, greater understanding of edge effects is important for management and conservation of communities (Ries et al. 2004, Hinchey et al. 2008).

Differences in food availability between the edge and interior of habitat patches can cause shifts in species abundances (Murcia 1995, Ries et al. 2004). Optimal foraging theory predicts that species will forage at locations providing the greatest benefit at the lowest energetic cost (Begon et al. 1996). If edges provide greater food availability, without any costs such as predation or increased energy use, it would be expected that abundances would be greater at the edge of patches. Fish diets can reflect food availability (Hettler 1989) and gut analysis has been used to assess diet variations across varying spatial and temporal scales (Hettler 1989, Fell et al. 1998, Hindell 2006).

Widebodied pipefish, Stigmatopora nigra, are an abundant seagrass fish species found across southern Australia (Gomon et al. 2008). Abundances of S. nigra are greater at seagrass edges than the middle of seagrass patches (Smith et al. 2008, Smith et al. 2010, Macreadie et al. 2009), which may be a result of greater food availability at the edge of patches (Smith et al. 2008, Macreadie et al. 2009). Stigmatopora nigra use a prehensile tail to grasp seagrass blades within the seagrass canopy where they are camouflaged from predators and feed on plankton in the water column (Howard & Koehn 1985, Jenkins & Sutherland 1997, Kendrick & Hyndes 2005).

The diet of Stigmatopora nigra consists mostly of small crustaceans, which can include up to 90.9 % planktonic copepods (Steffe et al. 1989, Kendrick & Hyndes 2005), suggesting that S. nigra depend on water movement for feeding opportunities. Seagrass

62 structure alters water movement, reducing flow across seagrass patches (Peterson et al. 2004). Decreasing water flow across seagrass beds may reduce the influx of planktonic copepods into the interior of patches, lowering food availability at the patch interior. Other prey items such as benthic crustaceans are also more abundant at seagrass edges than further into patches (Tanner 2005, Warry et al. 2009). Stigmatopora nigra may be responding to a resource (copepod abundance) that is itself responding to patch edges, known as a ‘cascading edge effects’ (Ries, et al. 2004). Similarly, seagrass structure can alter abundances of prey items such as harpacticoid copepods, amphipods and ostracods (Connolly 1995, Jenkins & Sutherland 1997, Jenkins et al. 2002), prey aggregations (Flynn & Ritz 1999) and syngnathid foraging (Ryer 1988, Flynn & Ritz 1999), which may cause variation in S. nigra diet and may partially explain why S. nigra prefer long dense seagrass compared to short sparse seagrass (Steffe et al. 1989).

Temporal changes in prey availability can affect species abundances (Begon et al. 1996). Stigmatopora nigra distributions may change according to copepod abundances which can undergo both seasonal (Jenkins & Hamer 2001) and diurnal (Walters & Bell 1986) variations. Stigmatopora nigra are visual predators (Howard & Koehn 1985) and may feed more effectively during the day than night; this could affect their dietary composition.

Abundances of Stigmatopora nigra collected in a previous study (Smith et al. 2010) , were greater at the seagrass edge than the middle during the day in autumn, but not at any other time. I hypothesised that if food availability was causing edge effects for S. nigra; abundances of prey in S. nigra guts sampled in Smith et al. (2010) would be greater at the seagrass edge than middle in autumn, but not at any other time, assuming diet reflects food availability (Hettler 1989). By determining the gut contents of S. nigra this study aims to (1) assess any differences in feeding between the edge and middle of seagrass patches, (2) assess if feeding patterns change over diel cycles or seasons, (3) assess changes in feeding patterns according to seagrass structure, and (4) determine differences in feeding patterns between S. nigra males, females and juveniles.

63 Methods

Field Sampling Fish were sampled from seagrass patches at Blairgowrie on the south-eastern coast of Port Phillip Bay, Victoria (38° 21’ 46” S, 144° 47’ 21” E). At Blairgowrie, the seagrass Heterozostera nigricaulis grows in patches of varying size in shallow (< 1 m) water. Gut analysis was done on a subsample of Stigmatopora nigra collected by Smith et al. (2010) who investigated the importance of patch size on the within patch distribution of fishes. Briefly, S. nigra were taken using a push net (0.5 x 1 m, with 1 mm mesh) in the seagrass within one metre of the seagrass-sand interface (edge) and at the middle of each of six patches. Samples were taken on four days and four nights during spring and autumn in 2006. Once collected, fish were anesthetised and preserved in ethanol for analysis in the laboratory. After sampling of fish was completed in each season, four seagrass samples were taken at both positions in each patch to determine seagrass density. Seagrass samples were collected by cutting all seagrass within a 0.25 cm2 quadrat, and preserving it for analysis in the laboratory.

Laboratory Methods Under a dissecting microscope, pipefish guts were removed, opened and contents emptied onto a microscope slide containing glycerol. Dietary items within the stomaches were identified to the lowest taxonomic group possible, counted to ascertain the abundance of prey at each position, and recorded. Copepod prey were divided into two groups: the generally planktonic calanoid and cyclopoid copepods were grouped together (planktonic copepods), and the generally benthic harpacticoid copepods were grouped together (harpacticoid copepods). The length of all fish was measured and fish were classified as male, female or juvenile. Although Stigmatopora nigra are sexually dimorphic, it is often difficult to distinguish between females and juveniles because they both lack pouches. I assumed all fish without a pouch larger than the smallest male were female (Steffe et al. 1989). Seagrass density estimates were taken from data used in Smith et al. (2010).

64 Data Analysis Data were checked for normality and homogeneity of variance using box plots and log transformed where required and reassessed (Quinn & Keough 2002). Differences in prey abundance in Stigmatopora nigra guts at the edge and middle of seagrass patches, were assessed using a repeated measures ANOVA design. Season (Autumn, Spring) and Time of Day (Day, Night) were treated as fixed factors; Position (Edge, Middle) was treated as a repeated measure. Abundances of total prey, harpacticoid copepod prey, planktonic copepod prey and Stigmatopora nigra length were used in analyses for all S. nigra sampled, and, male, female and juvenile S. nigra. Gut contents were averaged over patches to ensure there were no missing cells where fish were not sampled.

Analysis of seagrass density between positions and season was done using a repeated measures model using Season as a fixed factor and Position as the repeated measure, with seagrass patches used as replicates. Additionally, the relationship between Stigmatopora nigra gut content, and both S. nigra length, and, seagrass density, was assessed separately using linear regression analyses for total fish sampled, and males, females and juveniles gut content (Total prey, benthic prey, planktonic prey). By doing large numbers of regression analyses (28), I acknowledge that at least one significant result could arise by chance. Abundances were averaged over sampling days (n = patch 6  position 2  season 2  time of day 2 = 48) and treated as dependant variables in regression analyses.

Results

The guts of 350 Stigmatopora nigra were analysed; 179 from the edge of seagrass patches and 171 from the middle, including 120 males, 175 females and 55 juveniles. Copepods were the most common prey in guts (Table 11), with harpacticoid and planktonic copepods present in 89.4 % and 89.1 % of guts, respectively. Planktonic copepods were the most abundant prey observed in S. nigra guts averaging 70.4 prey items per gut. Other prey items included amphipods, tanaids, ostracods and isopods (Table 11). As only 12 juveniles were sampled in spring, season was removed as a factor from repeated measures analysis of juveniles. Repeated measures analyses of males (n = 15), females (n = 14) and juveniles (n = 5) had uneven replication because

65 fish were not collected within each position on every sampling occasion. Type III sum of squares, which are not influenced by the sample size in each cell (Quinn & Keough 2002) were used in analyses with uneven replicates.

Table 11. Mean abundance of each prey type per gut (±1 SE) from Stigmatopora nigra guts and frequency of occurrence. Data pooled across position, season and time of day.

Average Prey type abundance SE Frequency Total Prey 87.866 5.188 96.3 Planktonic Copepod 70.406 4.968 89.1 Harpacticoid Copepod 14.774 1.641 89.4 Gammarid Amphipods 1.811 0.186 56.9 Caprellid Amphipods 0.051 0.016 3.7 Ostracods 0.186 0.042 9.4 Gastropods 0.009 0.005 0.9 Isopods 0.066 0.024 3.4 Shrimp Larvae 0.111 0.032 5.7 Unidentified Worms 0.449 0.182 12.9 Crab Larvae 0.003 0.003 0.3 Tanaids 0.003 0.003 0.3 Cumaceans 0.003 0.003 0.3 Unidentified Prey 0.023 0.009 2.0

Position, season and time of day The abundance of prey in Stigmatopora nigra guts did not vary statistically between the edge and middle of seagrass patches, time of day or season for total prey, harpacticoid or planktonic copepods, except that harpacticoid copepods were more abundant in spring than autumn (Table 11, Figure 11). Male, female and juvenile S. nigra guts showed no difference in prey abundance between seasons, time of day or position, for total prey, harpacticoid copepods or planktonic copepods, with the exception of female S. nigra guts, where harpacticoid copepod abundance was greater in spring than autumn (Tables 12, 13, 14, 15, Figure 12).

66 6 7 Table 12. Results of repeated measure analysis of variance, assessing Stigmatopora nigra length, total prey abundance, harpacticoid copepods and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05).

Harpacticoid Planktonic Between Subjects Length Total prey copepods copepods Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Season 1 24.915 <0.001 0.372 0.553 7.1 0.021 2.754 0.123 Time of Day (TOD) 1 0.514 0.487 2.091 0.174 0.973 0.343 1.883 0.195 Season x TOD 1 0.864 0.371 1.464 0.25 0.086 0.775 0.51 0.489 Error 12

Within Subjects Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Position 1 5.905 0.032 0.232 0.638 0.706 0.417 0.671 0.429 Position x Season 1 0.203 0.66 0.576 0.463 1.404 0.259 0.001 0.98 Position x TOD 1 0.097 0.761 0.155 0.701 1.048 0.326 4.048 0.067 Position x Season x TOD 1 0.033 0.858 0.175 0.683 0.162 0.694 0.436 0.521 Error 12 90 a. b.

) 85 m m

( 80

h t g

n 75 e L 70

65 Edge Middle 30 c. d Position o

p 25 e p e o c 20 c n

a d i d o n 15 c u i t b c A

a 10 p r a 5 H 0 Autumn Spring Season Figure 11. Mean (± 1 SE) Stigmatopora nigra length during (a) autumn and spring, and (b) at the edge and middle of seagrass patches, and (c) mean (± 1 SE) S. nigra harpacticoid copepod gut abundance during autumn and spring.

68 6 9

Table 13. Results of repeated measure analysis of variance, assessing male Stigmatopora nigra length, total prey abundance, harpacticoid copepods and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05). Harpacticoid Planktonic Between Subjects Length Total prey copepods copepods Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Season 1 7.870 0.017 1.839 0.202 1.67 0.223 4.139 0.067 Time of Day (TOD) 1 1.397 0.262 0.261 0.620 0.057 0.816 0.542 0.477 Season x TOD 1 0.098 0.761 2.45 0.146 0.744 0.407 1.352 0.27 Error 11

Within Subjects Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Position 1 0.981 0.343 0.073 0.791 1.545 0.24 0.066 0.803 Position x Season 1 1.185 0.300 1.234 0.290 1.233 0.291 0.247 0.629 Position x TOD 1 0.116 0.740 0.097 0.761 1.336 0.272 0.308 0.590 Position x Season x TOD 1 0.003 0.957 0.140 0.715 0.712 0.417 0.153 0.703 Error 11 Table 14. Results of repeated measure analysis of variance, assessing female Stigmatopora nigra length, total prey abundance, harpacticoid copepods and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05). Harpacticoid Planktonic Between subjects Length Total prey copepods copepods Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Season 1 76.424 <0.001 0.012 0.916 6.657 0.030 0.694 0.426 Time of Day (TOD) 1 0.469 0.511 0.949 0.355 0.075 0.790 1.882 0.203 Season x TOD 1 0.946 0.356 0.166 0.694 0.023 0.883 0.021 0.889 Error 9

Within subjects Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Position 1 0.246 0.632 2.955 0.120 0.239 0.637 1.164 0.309 Position x Season 1 0.125 0.732 0.037 0.852 0.073 0.793 0.116 0.741 Position x TOD 1 0.125 0.732 0.032 0.861 3.189 0.108 0.969 0.351 Position x Season x TOD 1 0.001 0.978 0.625 0.450 2.346 0.160 0.002 0.969 Error 9 7 0 7 1

Table 15. Results of repeated measure analysis of variance, assessing juvenile Stigmatopora nigra length, total prey abundance, harpacticoid copepods and planktonic copepods between positions (edge, middle), seasons (autumn, spring) and time of day (day, night). Significant results in bold (p < 0.05). Harpacticoid Planktonic Between Subjects Length Total prey copepods copepods Source df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Time of Day (TOD) 1 0.348 0.597 3.075 0.178 1.645 0.290 3.022 0.181 Error 3

Within Subjects df F-ratio p-value F-ratio p-value F-ratio p-value F-ratio p-value Position 1 3.430 0.161 0.018 0.903 0.886 0.416 0.010 0.927 Position x TOD 1 9.170 0.056 0.017 0.905 0.021 0.893 0.020 0.895 Error 3 Stigmatopora nigra length Stigmatopora nigra lengths were significantly longer in spring than autumn and at the edge than the middle of seagrass patches (Table 12, Figure 11). Both male and female S. nigra were longer in spring than autumn (Table 13, 14, Figure 12). Regression analysis comparing S. nigra length and prey items showed that abundances of harpacticoid 2 copepod prey items increased with S. nigra size (df1,43 F = 8.24, p = 0.006, r = 0.161, 2 Figure 13) but there was no relationship with total (df1,43 F = 0.01, p = 0.941, r < 2 0.001), planktonic copepod prey abundances (df1,43 F = 1.22, p = 0.276, r = 0.028, 2 Figure 13) or seagrass density (df1,43 F = 2.20, p = 0.146, r = 0.049),. Male S. nigra 2 length was negatively related to total prey (df1,34 F = 6.41, p = 0.016, r = 0.159) and 2 planktonic prey (df1,34 F = 0.61, p = 0.039, r = 0.119) abundances (Figure 13) but there 2 was no difference for harpacticoid copepods (df1,34 F = 0.17, p = 0.682, r = 0.005) or 2 seagrass density (df1,34 F = 1.05, p = 0.314, r = 0.0.03). Female S. nigra length was 2 positively related to harpacticoid copepod abundance (df1,37 F = 5.21, p = 0.028, r = 2 0.123, Figure 13) and seagrass density (df1,37 F = 11.43, p = 0.002, r = 0.236, Figure 2 13), but there was no relationship with total prey (df1,37 F = 0.65, p = 0.425, r = 0.017) 2 or planktonic copepods (df1,37 F = 2.72, p = 0.108, r = 0.068). Juvenile S. nigra length 2 did not show any relationship with total prey (df1,14 F = 1.38, p = 0.260, r = 0.090), 2 = harpacticoid (df1,14 F = 0.29, p = 0.598, r 0.020) or planktonic (df1,14 F = 2.14, p = 2 2 0.166, r = 0.133) copepods or seagrass density (df1,14 F = 3.74, p = 0.073, r = 0.211).

72 100 Male Stigmatopora nigra

) 90 m m (

h

t 80 g n e L 70

60 100 Female Stigmatopora nigra

) 90 m m (

h t 80 g n e L 70

60 50 Female Stigmatopora nigra s t

u d g

i 40

o n i c

i t e c

c 30 a n p a r d a

n 20 H u b a 10

0 Autumn Spring Season

Figure 12. Mean (± 1 SE) male and female Stigmatopora nigra length, and female harpacticoid copepod prey abundance during autumn and spring pooling across positions, patches and time of day.

Seagrass density Seagrass density was significantly correlated with Stigmatopora nigra gut contents. As seagrass density increased the abundance of harpacticoid copepod prey increased (df1,43 2 F = 4.69, p = 0.036, r = 0.098, Figure 14), but planktonic copepods decreased (df1,43 F = 7.16, p = 0.011, r2 = 0.143, Figure 14). There was no relationship between seagrass

73 2 density and total prey items (df1,43 F = 1.86, p = 0.179, r = 0.042). Abundances of planktonic copepods in male S. nigra decreased with seagrass density (df1,34 F = 6.07, p 2 = 0.019, r = 0.151, Figure 14), but showed no relationship with total prey (df1,34 F = 2 2 1.52, p = 0.226, r = 0.043) or harpacticoid copepod prey (df1,34 F = 2.86, p = 0.100, r = 0.278). There was no relationship between seagrass density and the abundance of prey in the guts of female or juvenile S. nigra (Female, total prey, df1,37 F = 0.14, p = 0.713, 2 2 r = 0.004, planktonic copepods, df1,37 F = 1.74, p = 0.196, r = 0.045, harpacticoid 2 copepods df1,37 F = 3.95, p = 0.054, r = 0.096; Juvenile total prey, df1,14 F = 1.45, p = 2 2 0.2.49, r = 0.094, planktonic prey, df1,14 F = 2.04, p = 0.176, r = 0.176, harpacticoid 2 copepods df1,14 F = 3.91, p = 0.068, r = 0.218). Seagrass density did not vary statistically between the edge and middle of patches, and was marginally non-significant between seasons (Table 16).

Table 16. Results of repeated measures analysis of variance comparing seagrass density among positions and season. Significant results in bold (p < 0.05). Between Subjects Seagrass density df F-ratio p-value Season 1 4.757 0.054 Error 10

Within Subjects df F-ratio p-value Position 1 1.621 0.232 Season x Position 1 2.204 0.168 Error 10

74 2.0 3 a. d. e t d i c n o n

1.5 e c t a i t n d

c 2 o n a c u

p t b r u a a 1.0

g h

d

l t o a u t p o g e t 1

p g g

o 0.5 o o c L L

0.0 0 1.6 1.7 1.8 1.9 2.0 2.1 2.5 3 b. e.

e d i 2.0 c c o i n c i n a t o d c

t 2 e

n a k c 1.5 d u p n n o r b a a p l a a d e

p h

n p d t t u o o u

u 1.0 b c p g g

a e 1 g g p o o o L L c 0.5

0.0 0 1.85 1.90 1.95 2.00 4.4 c.

y Log Stigmatopora t i

s nigra length n e d

s 4.2 s a r g a e s

g 4.0 o L

3.8 1.8 1.8 1.9 1.9 2.0 2.0 0 5 Log0 Stigmatopora5 0 5 nigra length Figure 13. Relationship between mean (a) Stigmatopora nigra lengths across sampling days and mean harpacticoid copepod abundance, (b) female S. nigra length and harpacticoid copepod abundance, (c) female S. nigra length and seagrass density, (d) male S. nigra lengths and total prey abundance and (e) male S. nigra length and planktonic copepod abundance.

75 2.0 a.

e d i c o n c i

a 1.5 t d c n a u p r b a a

h 1.0

d t o u p g

e g p o o 0.5 L c

3 b.

e c c i n n a o t

d 2 k n n u a b l a p

t d u o g p 1 e g p o o L c

6 c.

e c c i n n a o t d k n 4 n u a b l a p

t d u o g p

e g 2 p o o L c

0 3.8 3.9 4.0 4.1 4.2 4.3 4.4 Log seagrass density

Figure 14. Relationship between mean seagrass density and mean Stigmatopora nigra (a) harpacticoid and (b) planktonic copepod prey, and (c) male S. nigra planktonic copepod prey.

76 Discussion

Stigmatopora nigra consumed a variety of prey at the edge and middle of seagrass patches. Stigmatopora nigra prey items in this study were similar to those reported in other studies where copepods were the main food source of S. nigra, representing between 85 % and 97 % of gut volume (Steffe et al. 1989, Kendrick & Hyndes 2005).

Greater food availability has been suggested to explain greater Stigmatopora nigra abundance at the edge of seagrass patches (Smith et al. 2008, Macreadie et al. 2009). Copepod abundances decline from the edge to the middle of seagrass patches (Warry et al. 2009, Macreadie et al. In press) which may be related to greater water movement at the edge of patches. Baffling of water currents caused by seagrass structure (Peterson et al. 2004) may reduce the influx of planktonic copepods to the patch interior. Optimal foraging theory predicts greater abundances of planktonic copepods at the edge of seagrass patches would increase abundances of S. nigra at patch edges. Previously Smith et al. (2010) found abundances of S. nigra were greater at the edge than middle of seagrass patches during the day in autumn. Using consumed prey items as a proxy for food availability, I hypothesised that if food availability caused S. nigra to be more abundant at the seagrass edge, prey items from S. nigra guts sampled at the seagrass edge during the day in autumn would be greater than those sampled at the middle, and there would be no difference during any other time. This hypothesis was not supported by this study. No differences were found between gut contents at the edge and middle of patches, suggesting that food availability is not causing edge effects for S. nigra.

Recent work (Macreadie et al. In press) has demonstrated food distribution to be an important determinate of edge effects in the distribution of Stigmatopora argus, a closely related relative of S. nigra. Several factors could explain the lack of difference in S. nigra diets between the edge and the middle of the patch at Blairgowrie. Seagrass patches at Blairgowrie range in size (Smith et al. 2010). In small patches there may not be any differentiation between edge and interior habitats in water movement and subsequent prey availability. Abundances of copepods and other S. nigra prey however, have been shown to decrease at only short distance into seagrass patches (> 5 m Tanner 2005, Warry et al. 2009), suggesting that even in small patches there will be some

77 decrease in prey availability at the patch interior. Seagrass density in Warry et al. (2009) however was much greater than at Blairgowrie, which may have had a greater affect in water baffling and subsequent copepod distribution. Alternatively, S. nigra feeding efficiency may decrease at the patch edge. Stigmatopora nigra are ambush predators (Howard & Koehn 1985) and the seagrass interior may provide better camouflage and reduce copepod movement, making copepod capture easier. Foraging at the edge may also present danger from predators inhabiting habitats adjacent to seagrass habitats. Finally, there may have been no difference in S. nigra gut content between the edge and middle of patches if S. nigra were moving between the edge and middle of patches.

Variation in fish lengths can affect the amount of prey eaten (Edgar & Shaw 1995b). In this study there was no relationship between Stigmatopora nigra length and abundance of either total or planktonic prey. Harpacticoid copepods, however, were more abundant in S. nigra guts with increasing S. nigra length and a similar pattern was shown in females. Similar to other pipefish species such as Urocampus carinirostris, overall and female S. nigra lengths, were greater in spring (Howard & Koehn 1985), which may be related to recruitment to seagrass patches during autumn. Greater S. nigra lengths in spring coincided with greater abundances of harpacticoid copepods in S. nigra guts, which may explain the relationship between length and harpacticoid abundance. Stigmatopora nigra lengths were greater in the middle of seagrass patches than the edge, but were unlikely to be masking any differences in diet because there was no relationship between length and gut content, and size differences between positions were only small (4.1 mm).

Male Stigmatopora nigra total gut prey abundance and planktonic copepod prey abundance decreased with greater length. Male gut fullness has previously been shown to be lower than that of females and is thought to be related to reduced foraging ability of males while brooding (Steffe et al. 1989). Reduced feeding abilities of larger brooding males may explain why prey abundances are lower for larger males then smaller ones.

78 Seagrass structure can affect Stigmatopora nigra abundances (Steffe et al. 1989) but are unrelated to S. nigra distribution between the edge and middle of patches (Smith et al. 2008, Smith et al. 2010). My findings support Steffe et al. (1989) who found that seagrass density did not affect total gut content, however, seagrass density may affect what prey S. nigra eat. Changes in seagrass density were correlated with changes in S. nigra gut content. Abundances of benthic harpacticoid copepods in gut samples increased as seagrass density increased, while abundances of planktonic copepods decreased as seagrass length increased. The relationship between seagrass density and S. nigra prey items could be attributed to prey availability. Harpacticoid abundance increases with seagrass density (Jenkins et al. 2002), explaining greater harpacticoid abundances in S. nigra guts in dense seagrass and greater planktonic copepod prey in sparse seagrass. Alternatively, changes in seagrass density could affect the foraging efficiency of S. nigra. Syngnathid foraging behaviour can change according to seagrass structure (Ryer 1988, Flynn & Ritz 1999) which may facilitate harpacticoid copepod capture in dense seagrass and restrict planktonic copepod capture. Seagrass structure, by changing prey availability or foraging behaviour, is therefore probably more important than seagrass edges in determining S. nigra prey types.

Stigmatopora nigra are visual predators (Howard & Koehn 1985), and therefore it would be expected that less prey are consumed at night when foraging is more difficult. In this study however, I found no differences in the amount of prey eaten between diel periods. The lack of any diel feeding pattern may be related to gut evacuation rates. Although there is no information relating to gut evacuation rates of syngnathids, if rates were high (i.e. ten hours to digest prey), fish sampled at night may still contain prey consumed during the day, confounding any pattern. Alternatively, many copepods are more active in the water column at night (Walters & Bell 1986), making S. nigra less reliant on vision to capture prey and therefore nullifying any expected diel gut differences.

Fish diets can reflect food availability (Hettler 1989), and changes in harpacticoid and planktonic copepod abundances relative to seagrass density suggest this is true. It should be acknowledged, however, that food availability may not be reflected in diets,

79 potentially explaining the lack of pattern in diets between the edge and middle. For instance, if prey densities are greater at the edge than middle, but dietary needs can be fulfilled at each position, then greater prey abundances at the edge will not be reflected in diets.

Changes in food availability across habitat patches can cause edge effects, or, are a reflection of edge mediated effects (Murcia 1995, Ries et al. 2004). However, in seagrass patches, Stigmatopora nigra gut analyses showed little difference in prey consumption between the edge and middle of patches, and prey consumption was more closely related to changes in seagrass structure. The lack of any difference in prey consumed by S. nigra at the edge and middle of patches, even when S. nigra abundances were higher, suggests that food availability is not causing abundances of S. nigra to be greater at the edge than the middle of seagrass patches. Instead, some other factor such as predation by larger piscivorus fish in the middle of patches may be determining within patch distributions of S. nigra.

80 5. Edge effects in patchy seagrass landscapes: the role of predation in determining fish distribution

ABSTRACT: Predation is often described as an underlying mechanism to explain edge effects. I assessed the importance of predation in determining edge effects in seagrass using two approaches: a video survey to sample predators at scales appropriate for seagrass edges, and a tethering experiment to determine if predation was an underlying mechanism causing edge effects. Underwater videos were placed at four positions: middle of seagrass patches; edge of seagrass; sand immediately adjacent to seagrass and sand distant from seagrass. Fish abundances and the time fish spent in view were measured. The main predatory fish (Australian salmon, Arripis spp.) spent more time over adjacent sand than other positions, while potential prey species (King George whiting, Sillaginodes punctata, recruits) were more common in the middle of seagrass patches. Other species, including the smooth toadfish, Tetractenos glaber, and whiting adults, spent more time over sand adjacent to seagrass than distant sand, which may be related to feeding opportunities. Whiting recruits and pipefish (Stigmatopora spp.) were tethered at each of the four positions. More whiting recruits were preyed upon at outer than inner seagrass patches, and survival time was greater in the middle of shallow seagrass patches than other positions. Relatively few pipefish were preyed upon, but of those that were, survival time was lower over sand adjacent to seagrass than at the seagrass edge or middle. Video footage revealed that salmon were the dominant predators of both whiting recruits and pipefish. The distribution of predators and associated rate of predation can explain edge effects for some species (whiting) but other mechanisms, or combinations of mechanisms, are determining edge effects for other species (pipefish).

Keywords: predation, seagrass, fish, edge effects, video sampling, tethering

81 Introduction

Edge effects are a fundamental concept in landscape ecology, and, as habitats continue to fragment, an understanding of edge effects is increasingly important for management and conservation (Ries et al. 2004, Hinchey et al. 2008). Habitat edges can change species abundances and distributions by altering physical conditions, resource distribution and species interactions (Murcia 1995, Ries & Sisk 2004). Predation can be an underlying mechanism changing species abundances at edges by increasing encounter rates between species, lowering refuge value, or enhancing foraging potential (Murcia 1995, Ries & Sisk 2004).

Seagrass is a conspicuous nearshore habitat that supports a rich and diverse array of fauna including many predatory species (Jackson et al. 2001). However, seagrass is under threat from a range of anthropogenic and natural sources including eutrophication, dredging and storms (Duarte 2002), leading to habitat fragmentation and increasing relative amounts of patch edges (Horinouchi 2007b). The seagrass/sand interface forms a distinct structural boundary that affects water flow (Peterson et al. 2004), light levels (Carruthers & Walker 1997), and sedimentation (Scoffin 1970). Changes to structural properties at the edge of seagrass can affect species abundance by increasing planktonic food supply and recruitment (Eckerman 1987) or changing foraging and escape behaviour (Orth et al. 1984).

Small species of fish show responses to seagrass edges that may be related to predation from larger fish (Smith et al. 2008). However, detecting edge effects of predatory, often transient fish in seagrass patches requires effective sampling at small spatial scales and adequate sampling designs (Connolly & Hindell 2006). Although the relative effect of an edge can vary according to patch size (Macreadie et al. 2009, Smith et al. 2010), such patterns cannot be confidently detected without sampling both the edge and interior of a patch.

Underwater video surveys are a non-destructive means of sampling mobile fish without altering behaviour (Cappo et al. 2003), and sample transient fish at fine scales appropriate for seagrass edges studies. Analysis of video footage, however, can be

82 problematic, and suitable methods depend on the hypothesis being tested and the type of video used. Recent studies have used MaxN (the greatest fish abundance within a block of time) to represent abundance (Cappo et al. 2003). When tested against other measures, MaxN has been shown to accurately reflect fish and species densities (Willis et al. 2000), eliminating problems associated with fish re-entry into frames (Cappo et al. 2003). However, MaxN is a very conservative estimate of fish abundance, and is most effective when bait is used to attract fish and, therefore, large numbers of fish stay within the field of view. Baited designs are, however, unsuitable for fine-scale sampling where fish may be attracted from nearby locations. Alternative methods for analysing video footage more appropriate to unbaited, fine-scale sampling designs include, recording the time fish spend in the field of view, or taking an average of MaxN over a number of blocks to reduce the influence of large schools on MaxN.

The presence of predators does not necessarily equate to greater predation levels. Assessing the role of predation in determining species distributions can be difficult and may vary according to water depth, seagrass structure and diel cycles (Edgar & Shaw 1995a, Jackson et al. 2006b, Horinouchi 2007b, Gorman et al. 2009). Tethering provides an experimental method of assessing predation, and has been used in a variety of habitats including coral reefs (Danilowicz & Sale 1999, Chittaro et al. 2005), seagrass (Bologna & Heck 1999, Peterson et al. 2001, Hovel & Lipcius 2002, Haywood et al. 2003, Horinouchi 2007a, Gorman et al. 2009) and unvegetated mudflats (Peterson et al. 2001, Haywood et al. 2003). Organisms that have been used in tethering experiments include fish (Dahlgren & Eggleston 2000, Smith & Hindell 2005), crustaceans (Peterson et al. 2001, Haywood et al. 2003), plankton (Bullard & Hay 2002) and seagrass seeds (Orth et al. 2006b).

Although a useful means of testing predation rates, tethering has its limitations and may produce biases (Peterson & Black 1994, Aronson et al. 2001). Tethering experiments are unable to estimate natural predation rates, instead reflecting maximum predation rates (Peterson & Black 1994). Tethering may alter prey behaviour, potentially increasing predator attraction (Danilowicz & Sale 1999), or restrict escape responses (Haywood et al. 2003). Experimental treatments may interact with tethered individuals

83 (Peterson & Black 1994), increasing susceptibility of animals to predation in one treatment more than another, in ways which would not occur in untethered animals (Haywood et al. 2003, Mills et al. 2008). While acknowledging such deficiencies, tethering is nevertheless a useful means of studying relative predation rates.

This study aims to: (1) assess predator abundances in different positions across seagrass landscapes, and (2) determine if predator abundance is causing edge effects in small seagrass fish species using a tethering experiment.

Methods

Field site The survey was done at Blairgowrie (38° 21’ 46” S, 144° 47’ 21” E), on the southern side of Port Phillip Bay, Victoria, Australia between December 2007 and February 2008. Blairgowrie supports patches of the seagrass Heterozostera nigricaulis that range in size from less than a 1 m2 up to 3790 m2. Seagrass at Blairgowrie supports many small seagrass-associated species of fish (Smith et al. 2008) and large predator fish species (Hindell et al. 2001). Water quality at Blairgowrie is high because of its close proximity to the entrance of Port Phillip Bay.

Field sampling Video sampling Remote video sampling was used to assess the distribution of mobile fish species. Sampling was done using four remote underwater cameras (3 Cannon HV20, 1 Sony CCD-TR16E). Cameras were fitted with wide angled lenses (0.07 focal length) that provided an underwater field of view of > 4.0 m2. Each camera was placed inside an underwater housing and clamped to a post 40 cm above the sea floor. Videos were used without bait to avoid attracting fish from nearby positions. At night, lights attached to posts were placed 50 cm behind and to either side of the camera. White light can affect fish behaviour, potentially attracting or deterring fish from the video field of view (Widder et al. 2005). To reduce the effect of light on behaviour, we used red light, which cannot be seen by most fish (Bowmaker 1995), including Stigmatopora nigra (Mosk et al. 2007) a common seagrass species at Blairgowrie (Smith et al. 2008). The

84 lights presented a field of view of 1.5 metres, less than during the day, and thus restricting direct comparisons between night and day samples.

Sampling design Day samples During the day cameras were placed in one of four positions within one of twelve randomly chosen patches: seagrass edge - within 1 m of the seagrass-sand interface; seagrass middle - halfway across the seagrass patch; adjacent sand - sand within 1 m of the seagrass-sand interface; and distant sand - sand at a distance from the seagrass edge equal to the distance to the middle of the seagrass patch. On each sampling day, each position was sampled three times from one of 12 independent patches, no patch was sampled twice (i.e. three replicates of each position from different patches on each day). Cameras were retrieved after one hour. Sampling was done on five days.

Night sampling Night sampling was done in the same way as day sampling. However, the much higher degree of difficulty in setting up videos and limits on light batteries restricted the number of samples and sampling time. At night, therefore, cameras were retrieved after 30 min, and samples were taken in as many positions as batteries would allow on each sampling night. Sampling was done on 6 nights.

Analysis of video footage Video footage was assessed using two different methods, MaxN average and time in view (TiV) at each position. MaxN average was calculated by averaging the maximum number of fish recorded in five 30 sec blocks for each video (i.e. whenever a fish was recorded, all fish from that species in the next 30 secs were recorded. At the end of the footage, five 30 sec blocks where fish were recorded were randomly chosen and averaged). When fish were recorded in fewer than five 30 sec blocks, the number of fish was still divided by five. MaxN average prevents overestimation of abundances caused by sampling a few large schools of fish. Australian salmon (Arripis spp.) were recorded in 10 sec blocks because of the fast swimming nature of this species. MaxN average will be referred to as MaxN for the remainder of this paper. TiV at each position provided an

85 estimate of how each position was used by fish. TiV was recorded as the total time in seconds that at least one fish of a given species was in the frame. Two rules were established to ensure robust analysis; 1) if a fish left the screen and returned from the same direction within 10 sec it was deemed to be the same fish and was not recounted in MaxN; 2) if a fish is lost from view (i.e. hidden in seagrass) and does not reappear within 30 sec it was deemed to have left the sampling area.

Tethering experiment Study species Tethering experiments used two common seagrass-associated fish: King George whiting (Sillaginodes punctata) recruits and pipefish (Stigmatopora spp.). Greater densities of pipefish have been found at seagrass edges compared with the seagrass interior, and the species is regularly collected on sand adjacent to seagrass (Smith et al. 2008). King George whiting are a commercially important species that recruits to seagrass habitats at Blairgowrie (Jenkins & Hamer 2001).

Tethering technique Fish were collected from nearby seagrass patches using a 10 m seine net and tethered using a 40 cm length of 0.006 mm monofilament nylon line through the lower jaw (whiting) or around the tail (pipefish). Tethers were attached to a timing device similar to that used by Peterson et al. (2001). Briefly, fishing line was connected to a ‘trigger’ that activated a reed switch, starting a clock. Underwater video was used to document behaviour and predator interaction for 82 whiting and 77 pipefish tethering trials.

Experimental design Tethered fish were placed at each of the four positions within six replicate seagrass patches over four sampling days (n = 4  6  4 = 96) for each species. Seagrass patches occured in two distinct bands; a shallow inner band and a deeper outer band. Three patches were used in each band, and seagrass band (outer, inner) treated as a factor. Tethered fish were assessed after 30 min, and their survival time and status recorded (eaten, alive or dead).

86 In the laboratory, footage of eaten or missing fish was assessed and the fate of the fish recorded. If a fish was missing but the timer was not triggered, and there was no video footage, the fish was recorded as dead. Dead fish and fish taken without setting of timers were omitted from the analysis.

Seagrass and depth Water depth and seagrass length was measured to assess any relationship with predator distribution and fish survival time. Depth measurements to the nearest 5 cm were taken prior to, and at the completion of, each video and tethering sample. Seagrass length measurements to the nearest mm were taken from 5 seagrass blades at the completion of video and tethering samples at the seagrass middle and edge.

Data analysis Video survey TiV and MaxN were adjusted for time (min-1) to account for any video samples that did not record for 60 minutes (four replicates), and four replicates were omitted from the analysis because of poor visibility. All data were checked for normality and homogeneity of variance and transformed where appropriate (Quinn & Keough 2002). TiV and MaxN for the most abundant species were analysed using a two-factor ANOVA, where Position (Distant Sand, Adjacent Sand, Seagrass Edge, Seagrass Middle) was treated as a fixed factor and sampling day was treated as a random factor. This was followed by three planned comparisons: vegetated and unvegetated habitats; sand and adjacent sand; seagrass edge and seagrass middle. Type III Sum of Squares, was used to reduce problems associated with uneven sample sizes where replicates had been omitted (Quinn & Keough, 2002). Depth and seagrass length were averaged for each replicate and regression analyses were used to compare each of TiV and MaxN with water depth and seagrass length (for seagrass length, sand positions were excluded).

Tethering experiment Traditional presence/absence data from the tethering experiments were analysed using chi-square analysis, comparing positions, and seagrass bands in separate analyses.

87 Survival time was converted to a proportion of 30 min to provide a relative measure across positions, and was assessed for normality and homogeneity of variance, arcsine transformed where required, and analysed using ANOVA. Unreplicated split-plot analyses were used to analyse survival times and fish lengths for each species. Seagrass band (outer, inner) was treated as a between subjects factors, and patch was treated as a random factor nested within seagrass band. Position (distant sand, adjacent sand, seagrass edge, seagrass middle) was treated as a within subjects factor. Response variables were averaged over sampling days. Tukey tests were done where there was a significant position effect. Logistic regression analysis (Wald statistic) was used in a separate analysis to determine if fish survival (eaten or survived) was affected by fish length, water depth or seagrass length. Survival time could not be tested against fish length, water depth or seagrass length using linear regression because many fish survived the entire tethering period skewing the data (right censored) violating the assumptions of homogeneity of variance.

Results

Video survey Day samples Over five sampling days, nineteen fish species from seventeen families were recorded. Smooth toadfish, Tetractenos glaber was the most commonly recorded fish species (occurring in 46 replicates), with small schooling fish (atherinids, clupeids, 27 replicates), Australian salmon (26 replicates) and both juveniles (12 replicates) and adults (9 replicates) of King George whiting also common.

Australian salmon Australian salmon did not differ in abundance among positions, but TiV was significantly greater at the sand edge than distant sand, and was intermediate at both seagrass positions (Table 17, Figure 15). TiV and MaxN of salmon increased with water depth (TiV: F1, 54 = 8.6, p = 0.005, MaxN: F1, 54 = 9.7, p = 0.003, Figure 16). TiV decreased with increasing seagrass length (F1, 26 = 8.5, p = 0.007, Figure 16) but there was no relationship for MaxN (F1, 26 = 3.2, p = 0.087).

88 Smooth toadfish Smooth toadfish were closely associated with seagrass and adjacent sand. Smooth toadfish TiV was greater in vegetated positions than unvegetated positions, and at the adjacent sand than distant sand (Table 17, Figure 15). Planned comparisons showed that toadfish MaxN was greater at adjacent sand than distant sand (Table 17, Figure 15).

Regression analysis revealed that when water depth increased, toadfish TiV (F1, 54 =

12.2, p = 0.001) and MaxN (F1, 54 = 23.5, p < 0.001) increased (Figure 16), but there was no relationship with seagrass length (TiV: F1, 26 = 0.4, p = 0.522, MaxN: F1, 26 = 1.0, p = 0.336).

89 9 0

Table 17. Analysis of variance comparing the time spent and MaxN in each position for each species sampled with videos. Significant values in bold (p < 0.05).

Australian salmon Smooth toadfish Small schooling fish King George whiting recruit TiV MaxN TiV MaxN TiV MaxN TiV MaxN df F p F p F p F p F p F p F p F p Day 4 3.35 0.020 1.61 0.344 0.72 0.584 1.94 0.124 5.23 0.002 5.72 0.001 2.98 0.032 2.65 0.049 Position 3 2.62 0.319 1.30 0.319 9.65 0.002 2.54 0.106 5.89 0.010 5.62 0.012 1.85 0.192 1.67 0.225 SG E v SG M 1 <0.01 0.984 0.16 0.695 2.66 0.129 0.03 0.872 0.40 0.541 1.88 0.196 0.98 0.342 0.05 0.820 Adj Sand v Dist Sand 1 7.56 0.018 3.11 0.103 7.90 0.016 7.11 0.021 0.13 0.541 0.12 0.738 0.62 0.446 1.50 0.244 SG v Unveg 1 0.30 0.596 0.60 0.451 19.42 0.001 0.49 0.449 16.73 0.001 14.12 0.003 3.69 0.079 3.39 0.090 Position  Day 12 0.65 0.788 0.35 0.970 0.57 0.854 1.45 0.190 0.97 0.496 0.71 0.496 2.13 0.039 1.92 0.065 Error 36 2.5 25 Day 1 1

1 Australian King George whiting recruits -

- Day 2 n

n Day 3 2.0 i

i salmon 20 Day 4 m m .

. Day 5 V

V 1.5

i 15 i T T

n n 1.0 10 a a e e M M 0.5 5

0 14 1 Small schooling fish - Small schooling fish 1 - 12 n 4 i n i m .

m 10 .

N 3 V x i 8 a T

M

n 2 6 a n e 4 a e M 1 2 M 0 0 12

1 Smooth toadfish Smooth toadfish - 1

n 0.03 - 10 i n i m .

m 8 . N

x 0.02 V i a

T 6

M

n n

a 4

a 0.01 e e

M 2 M 0 0.00 0.6 Dist Sand Adj Sand SG Edge SG Mid King George whiting adult 1 - 0.5 n

i Position m

. 0.4 V i

T 0.3

n

a 0.2 e

M 0.1

Dist Sand Adj Sand SG Edge SG Mid Position

Figure 15. Mean (± 1 SE) TiV in each position for Australian salmon, smooth toadfish, small schooling fish and King George whiting adults, TiV at each position on each sampling day for King George whiting recruits, and mean MaxN (± 1 SE) at each position for small schooling fish and smooth toadfish.

Small schooling fish Small schooling fish TiV and MaxN were greater in seagrass than sand positions (Table 17, Figure 15). Small schooling fish TiV in each replicate increased with seagrass length (F1, 26 = 12.4, p = 0.002, Figure 16) but there was no significant relationship

91 between MaxN and seagrass length (F1, 26 = 3.6, p = 0.068) or either TiV (F1, 54 = 1.1, p

= 0.307) or MaxN (F1, 54 = 0.9, p = 0.346) and depth.

King George whiting King George whiting adults and recruits showed different patterns; adults were associated with adjacent sand and recruits with the seagrass middle. Adult whiting were recorded in only nine replicates, restricting analysis, but over 50 % of recordings (5) were in adjacent sand where TiV was much greater (Figure 15). Whiting recruit TiV showed an interaction between positions and days, being greater in the middle than other positions on the first two sampling days (Table 17, Figure 15). There was no significant difference for MaxN (Table 14). There was no relationship between whiting

recruits and either water depth (TiV: F1, 54 = 1.5, p = 0.225, MaxN; F1, 54 = 1.6, p =

0.216) or seagrass length (TiV: F1, 26 = 0.4, p = 0.847, MaxN; F1, 26 = 0.1, p = 0.805).

0.7 2 2.0 Australian salmon Australian salmon Australian salmon 1 - 1 1 n - - i n n i i

1.5 m

. 0.5 m m . . N x V V i i a T T 1.0

1 M t t

t 0.3 o o o o o r r o r h h t t 0.5 h t 4 4 4 0.1

0.0 0

5 0.3 3 Smooth Smooth toadfish Small schooling fish 1 1 1 - - - toadfish n n n

4 i i i m m m . . . 0.2 2 N V V i i

3 x T T a

t t M o o

t o o 2 r r o 1 o

0.1 h h t t r 4 4 h 1 t 4

0 0.0 0 0 50 100 150 0 50 100 150 0 10 20 30 Mean Depth (cm) Mean Depth (cm) Mean Seagrass Length (cm)

Figure 16. Relationship between each of TiV and MaxN and water depth for Australian salmon and smooth toadfish, and between TiV and seagrass density at the middle and edge of patches for Australian salmon and small schooling fish.

92 Night Samples There was little difference in fish distribution among positions at night. A total of 31 night samples were taken over six nights, nine at the distant sand, eight at the adjacent sand and seagrass middle but only six at the seagrass edge. Only seven species were recorded in night samples with small schooling fish (26 replicates) and smooth toadfish (19 replicates) the most common. Australian salmon were not seen at night. All positions were not represented on every sampling day, therefore there was no position  day interaction data, and consequently analysis was not done. However, TiV in each position and MaxN for both small schooling fish and toadfish showed little difference among positions (Figure 17).

35 0.4 Small schooling fish Small schooling fish 1 - 1 - n i n i 0.3 m

25 . m . N V x i a T

0.2 M n

15 a n e a e

M 0.1

5 M

1.8 0.018

Smooth toadfish 1 Smooth toadfish - 1 n - i n i m . m

. 1.2 0.012 N x V i a T

M

n n a

0.6 a

e 0.006 e M M

0.0 0.000 Dist Sand Adj Sand SG Edge SG Mid Dist Sand Adj Sand SG Edge SG Mid Position Position

Figure 17. Mean (± 1 SE) TiV and MaxN in each position during night samples for small schooling fish and smooth toadfish.

Tethering Experiment A total of 96 of each of whiting recruits and pipefish were tethered over 8 days. Of these, 47 whiting were eaten by predators. Only 15 pipefish were eaten, and none was eaten in the middle of seagrass patches. Additionally, 13 pipefish were removed from analysis because they were dead, escaped, or gone and the timing devices had not been

93 triggered. Tethered fish length was not different across positions or seagrass bands (Table 18) for either species.

Traditional analysis comparing eaten and surviving fish revealed few significant results, with no difference across positions for either whiting (χ2 = 0.631, df = 3, p = 0.889) or pipefish (χ2 = 6.274, df = 3, p = 0.067). Comparisons between seagrass bands, however, found more whiting were eaten at outer seagrass patches than inner patches (χ2 = 10.150, df = 1, p = 0.001), but there was no difference for pipefish (χ2 = 0.621, df = 1, p = 0.430).

Whiting survival time Whiting survival time was greatest in the middle of inner seagrass patches where only one fish was eaten, while in outer patches survival time was lowest in the middle of seagrass patches, but there was no overall difference among positions or seagrass bands (Table 18, Figure 18). Survival was not significantly related to water depth (Z = 1.48, p = 0.138), seagrass length was marginally non-significant (Z = -1.94, p = 0.053) but as whiting length increased, chance of survival decreased (Z = 2.62, p = 0.001, Figure 19).

Pipefish survival time Pipefish survival time was greatest at the middle of seagrass patches where no fish were eaten. Pipefish survival time was lower over adjacent sand than both the seagrass edge (p = 0.036) and middle (p = 0.012) but there was no difference between other positions or seagrass bands (Table 18, Figure 18). There was no relationship between pipefish survival and either water depth (Z = -0.66, p = 0.507) or pipefish length (Z = -0.20, p = 0.840). Logistic regression analysis was not done between pipefish survival and seagrass length because only four fish were preyed upon in seagrass.

94 Table 18. Analysis of variance comparing survival times of King George whiting recruits and pipefish at different positions and seagrass bands. Significant values in bold (p < 0.05).

Fish Length Survival Time King George King George whiting recruits Pipefish whiting recruits Pipefish Between Subjects Df F p F p F p F p Seagrass band 1 0.06 0.798 0.00 0.985 5.13 0.086 0.07 0.807 Patch (Seagrass band) 4

Within Subjects Position 3 0.48 0.703 0.03 0.992 0.68 0.582 4.72 0.021 Position x Seagrass band 3 1.67 0.227 3.45 0.052 4.12 0.032 0.52 0.677 Position x Patch (Seagrass band) 12 9 5 35 King George whiting recruits Outer patches 30 Inner patches

) 25 n i m (

e 20 m i T

l

a 15 v i v r u

S 10

5

0 30 Pipefish

25 )

n 20 i m (

e m

i 15 T

l a v i

v 10 r u S 5

0 Dist Sand Adj Sand SG Edge SG Mid Position

Figure 18. Mean (± 1 SE) survival time of King George whiting recruits across positions and seagrass bands, and pipefish across positions.

Predators Australian salmon were the most common predator of both whiting recruits and pipefish. Salmon accounted for 43 predation events on whiting, followed by silver gulls, Larus novaehollandiae (3 predation events) and flathead (Platycephalidae spp. 1 predation event). Salmon accounted for 14 of the 15 predation events on pipefish; the other predation event was by a shrimp from the family Callianassidae.

96 1.0

g n n e i t K a

0.8 f e

o

g y n t i i l e i b b

0.6 a g b n i o t r i p h 0.4 w d

e e t g c i r d o e e

r 0.2 G P

0.0 40 45 50 55 60 65 70 75 80 Length (mm)

Figure 19. Relationship between the probability of King George whiting recruits being eaten after 30 min and King George whiting length.

Discussion

Predation can influence the distribution of species in and around seagrass habitats. This study found that both predator and prey distribution and prey survival times during the day can change according to both position within a seagrass patch, and the location of that patch within the seagrass landscape. My findings support previous work suggesting predation at seagrass edges is greater than at other seagrass locations (Bologna & Heck 1999, Gorman et al. 2009). Edges can facilitate greater encounter rates between predators and prey, decreasing prey abundances at edges through greater mortality rates or predator avoidance (Ries et al. 2004). Australian salmon were the predominant predator of tethered whiting and pipefish and is a common predator of many small fish in seagrass (Robertson 1982, Hindell et al. 2000). Salmon spent more time patrolling sand adjacent to seagrass than other positions and can be linked to predation and distribution of King George whiting recruits, and pipefish predation. The presence of salmon over adjacent sand was directly related to lower pipefish survival times over adjacent sand. Likewise, survival times of whiting recruits were greater in the middle of nearshore patches, but not in deeper patches. Whiting recruits spent more time in the middle of patches on the initial two days of sampling, suggesting the presence of

97 salmon was affecting whiting distribution within patches. Whiting recruits undergo ontogenetic changes, moving from seagrass to sand habitats during January (Jenkins & Wheatley 1998), which may explain the lack of consistency in whiting TiV over days. Moving to sand habitats coincides with changes in colouration and may initiate schooling behaviour, making recruits less susceptible to predation from salmon, and therefore spending less time in the middle of patches.

Seagrass edges can provide a foraging area or barrier for transient predators (Horinouchi 2007b). Syngnathids and gobiids are common prey of salmon (Robertson 1982, Hindell et al. 2000), are more abundant at seagrass edges, and can “spill” onto sand adjacent to seagrass (Smith et al. 2008). The seagrass edge and adjacent sand therefore provide salmon with a favourable foraging area creating a positive edge response (Ries et al. 2004). Survival times of pipefish were lower on adjacent sand compared to other positions, and no fish were eaten from the middle of patches. Therefore, it would be expected that if predation was causing edge effects for pipefish, abundances would be greater in the middle of patches where predation is lower. However, pipefish abundances are greater at the edge of patches than the middle (Smith et al. 2008, Macreadie et al. 2009), suggesting that pipefish balance predation pressure with some other mechanism that maximises energy and growth at the edge (Orth et al. 1984). Planktonic food availability, through greater water flow at the seagrass edge, has been suggested to explain greater abundances of pipefish at the edge (Smith et al. 2008, Smith et al. 2010, Macreadie et al. 2009), and, although gut analysis has found little differences in prey consumption between the edge and middle, supplementary food experiments have indicated that food availability may influence pipefish abundances within seagrass patches (Macreadie et al. In press).

Seagrass provides small fish species with a refuge from predators (Jackson et al. 2001), and changes in seagrass structure and presence of predators can affect the refuge value of patches (Horinouchi 2007b). Survival time in seagrass positions was generally greater than sand positions, supporting the premise that the presence of seagrass structure provides a refuge from predators (Hindell et al. 2002). Survival time of whiting in deeper outer patches was the exception, where survival time was lower in

98 seagrass than sand. Predator abundances are thought to increase in deep water (Sheaves 2001), and I found salmon TiV and abundance increased with water depth. Seagrass may disrupt schooling behaviour (Flynn & Ritz 1999), but deeper water in outer patches may allow salmon to maintain schools within seagrass patches, making whiting more prone to predation in deeper beds. Seagrass density, however, did not influence survival of tethered fish, in contrast to studies on crustaceans that found dense seagrass enhances survival (Haywood & Pendrey 1996, Hovel & Fonseca 2005). The predator/prey relationship between salmon and small schooling fish may be influenced by seagrass density; as seagrass density increased, salmon TiV decreased and small schooling fish TiV increased. Foraging ability of chase and attack predators such as salmon can be affected by seagrass density (Flynn & Ritz 1999), creating a refuge for small schooling fish. Similarly, in the absence of salmon at night, small schooling fish were recorded in all positions, suggesting they will move onto sand when predators are not present.

Sand adjacent to seagrass patches was used more than distant sand by Australian salmon, smooth toadfish and adult King George whiting, supporting previous findings suggesting that unvegetated sand close to seagrass is more important than more distant sand (Ferrell & Bell 1991, Smith et al. 2008). Sand adjacent to seagrass might represent an enhanced habitat for seagrass fishes (Smith et al. 2008). Salmon, toadfish and adult whiting may use adjacent sand as an enhanced habitat because there is greater food availability than distant sand. Toadfish and adult whiting feed on similar prey over sand (Robertson 1980) that includes polychaetes and decapods which have shown patterns of greater abundances on sand close to seagrass than more distant sand (Tanner 2005). Alternatively, toadfish, spent more time in seagrass than sand positions and may be ‘spilling over’ from the preferred seagrass habitat onto adjacent non-preferred sand habitat.

Remote underwater videos are a non-destructive means of sampling mobile fish species (Cappo et al. 2003) at spatially-explicit scales appropriate to investigating seagrass landscape effects. Assessment of video fish survey methods is well documented (Willis et al. 2000, Watson et al. 2005, Stobart et al. 2007), but there has been little comparison of video analysis techniques. MaxN is used in most baited video studies (Willis et al.

99 2000, Denny et al. 2004, Watson et al. 2005, Stobart et al. 2007) and correlates well with species richness and abundance (Willis & Babcock 2000). Other methods of analyses include time to first appearance (Cappo et al. 2004) and total fish counts (Laurel et al. 2007). I found differences between the two methods used in video analysis. In the case of Australian salmon, significant results were found for time in each position but not for MaxN, therefore the type of video analysis can influence the conclusions of the study. Baited video attracts large numbers of fish to the area, increasing variability across replicates. In this study, with the exception of the small schooling fish, the species were generally solitary (smooth toadfish, banjo sharks) or formed small schools (salmon, whiting), reducing any chance of variation between samples, and therefore, are unlikely to show variation in MaxN. Therefore, it is important to consider target species and the question that is being asked when deciding which method of video footage analysis to use.

Tethering experiments are an effective way of testing relative predation rates across locations and habitats, and, although they have been used extensively in recent studies (Bologna & Heck 1999, Smith & Hindell 2005, Horinouchi 2007a, Rypel et al. 2007, Mills et al. 2008), the limitations of tethering are well documented (Peterson & Black 1994, Aronson et al. 2001). By using not only tethering timers, but also video, I was able to assess survival time between habitats, predators preying on tethered fish, and any artefacts associated with tethering. Video footage showed tethering had little effect on the behaviour of whiting recruits compared to non-tethered whiting seen in video footage, except to restrict escape from predators (T. Smith pers obs). The high mortality rate of pipefish not associated with predation suggests that pipefish may not be a suitable species for tethering because of their poor swimming ability. Predation on pipefish by Callianassid shrimp highlights the artefacts that can be associated with tethering experiments. Callianassids are not a natural predator of pipefish but were able to take advantage of the restricted mobility of the tethered fish. However, with the use of video, and carefully selected prey species and sampling design, tethering is a useful means of testing predation across positions and habitats.

100 Habitat edges can change species interactions by altering movement patterns, inducing species mortality, increasing cross boundary subsidies and creating new interactions (Fagan et al. 1999). Predation is the most commonly cited interaction used to explain edge effects, where one organism benefits at the expense of another (Ries et al. 2004). In the case of Australian salmon (predator) and King George whiting recruits (prey) in shallow seagrass, the seagrass edge is a complementary habitat for salmon, but negative for the whiting prey, reducing abundances or forcing prey species to the interior of the patch where encounter rates with the predator are lower. Although the same predation climate would apply for pipefish, abundances of pipefish are greater at the seagrass edge (Smith et al. 2008), indicating that some other resource, probably food availability, is outweighing the risk of predation. The role of predation and species interactions in determining edge effects can be great for some species (whiting) or minimal for others (pipefish), where the distribution of other resources is of greater importance than the threat of predation. As fragmentation of seagrass and other habitats continues, it is becoming increasingly important to establish the effects that edges have on patch associated fauna, and the mechanisms underlying such effects.

101 6. General Discussion

General Discussion Edge effects are an important landscape component that can alter physical characteristics, flows and species composition (Murcia 1995, Ries et al. 2004). Fragmentation of habitats, particularly by human induced factors, increases edge habitat and is a threat to biodiversity (Fletcher et al. 2007). Therefore, understanding the role of edge effects on species distributions, and the underlying mechanisms that drive edge effects, is fundamental to conservation and management strategies (Ries et al. 2004, Fletcher et al. 2007). Habitat edges can affect species abundances by changing abiotic conditions, predator interactions and resource distributions (Murcia 1995, Fagan et al. 1999, Ries et al. 2004).

Seagrass habitats are naturally patchy and support a diverse range of fauna (Jackson et al. 2001), but increasing fragmentation through human disturbance is greatly increasing the amount of edge habitat (Hinchey et al. 2008). Seagrass edges alter water currents (Peterson et al. 2004), sedimentation (Scoffin 1970) and seagrass structure (Orth et al. 1984, Horinouchi 2007b) all of which can affect predation, food availability and recruitment (Orth et al. 1984, Tanner 2003, Horinouchi 2007b). The role of edge effects in seagrass however is equivocal, with few studies showing any response to seagrass edges (Connolly & Hindell 2006). In this thesis I have demonstrated not only that fish distribution responds to seagrass edges, but also that predation is one of the underlying causes of edge effects for some fish species.

The importance of edge effects to seagrass fishes Seagrass fish showed a strong response to habitat edges throughout this study. Previously, responses of fauna to seagrass edges have been few, with a review by Connolly & Hindell (2006) finding only four of 44 fish studied responding to seagrass edges. Within this study eight fish species responded to edges, double that from Connolly & Hindell’s (2006) study. Overall abundance of small permanent and cryptic fish, and of four individual species (Stigmatopora nigra, S. argus, Nesogobius maccullochi, Cristiceps australis) was higher at

102 the seagrass edge than the middle of patches, and over adjacent sand than distant sand. Underwater video, was used to sample fast-moving fish, finding that smaller species (Sillaginodes punctata recruits and small schooling fish) preferred the middle of patches, while larger fish (Arripis spp., adult S. punctata, Tetractenos glaber) spent more time over sand adjacent to seagrass patches. Seagrass structural characteristics (length, density, biomass, epiphyte biomass) did not explain greater abundances at the seaward edge of patches than the middle, but probably contributed to differences in abundances at the shoreward edge. Although unvegetated sand habitats are generally thought to be less important to fish than seagrass habitats (Orth et al. 1984, Ferrell & Bell 1991, Jenkins et al. 1997), I found that sand adjacent to seagrass patches was used extensively by seagrass fish in relation to more distant sand. The use of sand edges by seagrass fish suggests that fish are either “spilling over” onto the sand from seagrass, or, that sand provides some resource that is not found within seagrass. This thesis has contributed significantly to our knowledge of fish that respond to edges, however, there are several limitations within the study. Seagrass patches in this study were comparatively small compared to other studies (e.g. Jelbart et al. 2006) which may limit or enhance fish response to edges (see Chapter 3), and although seagrass structure and depth were taken into account throughout the thesis, their influence on fish responses was not experimentally tested. Future work incorporating experimental tests using ASUs of varying seagrass density at the edge and middle of large patches will help determine if seagrass structure or position within a patch is more important to fish. Such experiments will be important not only for seagrass landscape ecology, but for landscape ecology in general.

Patch and edge quality can influence how species respond to edges (Ries et al. 2004) but are difficult to control in both terrestrial and marine landscapes. Changes in seagrass structure, water currents, sediment composition or other variables may make some edges more valuable than others, but such characteristics have not yet been identified. Fish responses to not only varying edge quality, but also surrounding matrix quality will become clearer with manipulative experiments. Through the use of ASUs, predator exclusion cages, tethering and food supplementation, and taking into account current strength and direction,

103 sediment size and algae content, the quality of seagrass edges can be tested. Experimental studies on edge quality may explain some of the inconsistencies in both marine and terrestrial landscape ecology research. Identifying the most appealing edges has important ecological and management implications, and can be used to establish management strategies to conserve edges that are of the most importance.

Seagrass patch size affects the distribution of fish at patch edges Patch edge and area effects can interact, making it difficult to establish causes of within patch distributions (Fletcher et al. 2007). To differentiate between patch edge and area effects for fish in seagrass habitats I sampled the seagrass edge, 2 m into a patch and the middle of different sized patches. Two species, Stigmatopora nigra and Nesogobius maccullochi, displayed abundances at the edge and middle that were affected by the size of the patch. In small patches there was little difference between abundances at the edge and middle but in large patches there was a positive edge effect for S. nigra and a negative effect for N. maccullochi. These results suggest that below a certain size patches become “all edge” and that edge effects may actually be driving previously reported area effects by changing the magnitude of the edge effect in different size patches. In a management scenario, determining when a patch becomes “small” or reduced to entirely edge habitat will become an important conservation tool for species that prefer the patch interior, but will require further study that incorporates sampling at a variety of distances into patches and at a greater variety of patch sizes. Changes in edge effects related to patch area have important implications for researchers and may explain some of the variability in studies on edge effects.

Mechanisms explaining edge effects The importance of food as a determinant of edge effects A variety of mechanisms have been suggested to explain edge effects in seagrass patches, the most common being food availability and predation (Connolly & Hindell 2006). To determine if food availability was causing edge effects for fish in seagrass habitats, gut analysis was done on Stigmatopora nigra. Stigmatopora nigra is one of the most common

104 seagrass species, and is more abundant at seagrass edges than the interior. There was little difference in prey abundances between fish sampled at the edge and middle of patches, although prey types eaten were affected by seagrass density. Lack of difference between gut content at the edge and middle of patches suggests that food availability is not causing edge effects in S. nigra populations. Factors such as fish movement, distribution and patch size may be masking any patterns in gut prey abundances. Recent work by Macreadie et al. (In press) has experimentally shown that food availability can cause shifts in pipefish abundance from the edge to middle of patches, and therefore food availability as a cause of edge effects should not be disregarded.

The importance of predation as a determinant of edge effects There was little evidence to suggest that food availability was causing edge effects in seagrass, therefore the role of predation was assessed. Greater predator abundance and predation rates at the seagrass edge or middle could be causing fish to inhabit areas with lower encounter rates with predators or greater refuge potential. Using underwater video I was able to sample fast-moving fish at fine scales, and assess abundance and use of the edge and middle of seagrass patches. Australian salmon (Arripis spp.) is a common piscivore in seagrass habitats, feeding on many of the small seagrass fish that displayed edge effects such as pipefish (Stigmatopora spp.), King George whiting (Sillaginodes punctata) recruits and Gobiidae. Salmon spent more time over sand adjacent to seagrass than distant sand, and were more prominent in deep than shallow water. Potential prey of salmon on the other hand, such as whiting recruits, were more common in the interior of patches. The presence of salmon at the edge of patches could therefore be causing whiting recruits to inhabit the interior of patches where they are less likely to encounter salmon. However, other small prey species such as pipefish and Nesogobius maccullochi had earlier been shown to inhabit the edge, suggesting that predation by salmon may not influence distribution. To assess if greater time spent over adjacent sand by salmon was the mechanism causing the distribution of prey species, a tethering experiment was conducted. Tethering whiting recruits and pipefish revealed that predation by salmon was causing edge effects for whiting but not for pipefish. Whiting survival was greatest in the middle of

105 shallow, inshore patches, where few salmon were recorded. Pipefish survival was also greatest in middle of patches, where no predation was recorded. However, pipefish abundances are greater at the seagrass edge, exposing them to greater predation risk from salmon, suggesting that pipefish are trading off greater predation risk at the edge of patches with some other factor. Predation could explain the distribution of whiting recruits within seagrass patches, but was unable to explain the greater abundances of pipefish at the habitat edge.

Seagrass edges play an important role in determining fish distributions within seagrass habitats. Strong and consistent edge effects were seen in a number of common seagrass fish species that could be attributed to a variety of mechanisms. Sillaginodes punctata recruits prefer the interior of patches, while Stigmatopora spp. and Nesogobius maccullochi are more abundant at the seagrass edge. Sand adjacent to seagrass is used extensively by a variety of small seagrass fish as well as larger fish, including predatory Arripis spp. and commercially important S. punctata adults. Patch size, predation, and seagrass structure and water depth between the shoreward edge and middle of patches, all contributed to differences in species abundances between the edge and middle of patches. Although the presence of edge effects and factors contributing to edge effects has been established, further investigation needs to be done to assess how the quality of seagrass edges and surrounding habitats affects fish assemblages at seagrass edges. In a larger landscape context, understanding how fish communities at seagrass edges are affected by the proximity to other seagrass patches, alternate fish habitats, and corridors may explain inconsistencies among seagrass edge studies, and provide better strategies to governing nearshore landscape management. The generality of results across differing temporal and spatial scales also needs to be examined to determine if edge effects in Port Phillip Bay are relevant to other locations and are consistent across years.

Habitat loss and fragmentation increases the relative importance of habitat edges and will become increasingly important into the future. This thesis provides an insight into some of the potential consequences of seagrass fragmentation on seagrass fish and ecosystems. My

106 results suggest that patchy seagrass habitats that have high edge and adjacent sand ratios provide optimal habitats for many fish species. Habitat fragmentation may, therefore appear beneficial in seagrass landscapes. Fragmentation, and associated habitat loss, would have a detrimental effect on seagrass ecosystems, even on species that utilise the edge, particularly as core habitat is lost as patches become smaller. For example, I have shown that loss of seagrass interior habitat and increased edges as patches become smaller will make juvenile S. punctatus and small schooling fish more susceptible to predation from Arripis spp. Reduction in preferred habitat and greater predation of juvenile S. punctata could impact on adult populations of S. punctata, an important commercial and recreational fishing species that uses the sand associated with edge extensively. Prevention of further seagrass loss, and effective management of seagrass habitats in Port Phillip Bay, is therefore important to ensuring the ongoing sustainability of S. punctata populations. Further investigation is needed to establish when the benefits of fragmentation are outweighed by overall habitat loss. Small patches can effectively act as edge habitat only, so that as patches become increasingly smaller, there will be a point where the patch cannot sustain the population density of larger patches. Knowing at what patch size the ratio between edge and area becomes detrimental to edge associated species is important information for managers of seagrass ecosystems.

107 7. Bibliography

Anderson T (2003) The functional relationship between temperate fishes and the associated seagrass landscapes. PhD, The University of Melbourne Aronson RB, Keck KL, Valentine JF (2001) Measuring predation with tethering experiments. Mar Ecol Prog Ser 214:311-312 Ball D, Parry G, D., Heislers S, Blake S, Werner G (2006) Analysis of Victorian seagrass health at a multiregional level. Report No. 83, Department of Primary Industries Victoria, Queenscliff Barbosa O, Marquet PA (2002) Effects of forest fragmentation on the beetle assemblage at the relict forest of Fray Jorge, Chile. Oecologia 132:296-306 Begon M, Harper JL, Townsend CR (1996) Ecology, Blackwell science, Oxford Bell JD, Westoby M (1986) Abundance of macrofauna in dense seagrass Is due to habitat preference, not predation. Oecologia 68:205-209 Bell SS, Brooks RA, Robbins BD, Fonseca MS, Hall MO (2001) Faunal response to fragmentation in seagrass habitats: implications for seagrass conservation. Biol Conserv 100:115-123 Bender DJ, Contreras TA, Fahrig L (1998) Habitat loss and population decline: a meta- analysis of the patch size effect. Ecology 79:517-533 Berkenbusch K, Rowden AA (2007) An examination of the spatial and temporal generality of the influence of ecosystem engineers on hte composition of associated assemblages. Aquatic Botany 41:129-147 Bloomfield AL, Gillanders BM (2005) Fish and invertebrate assemblages in seagrass, , saltmarsh and nonvegetated habitats. Estuaries 28:63-77 Bologna PAX (2006) Assessing within habitat variability in plant demography, faunal density, and secondary production in an eelgrass (Zostera marina L.) bed. J Exp Mar Biol Ecol 329:122-134 Bologna PAX, Heck KL (1999) Differential predation and growth rates of bay scallops within seagrass habitats. J Exp Mar Biol Ecol 239:299-314

108 Bologna PAX, Heck KL (2002) Impact of habitat edges on density and secondary production of seagrass-associated fauna. Estuaries 25:1033-1044 Boström C, Jackson EL, Simenstad CA (2006) Seagrass landscapes and their effects on associated fauna: a review. Est Coast Shelf Sci 68:383-403 Bowden DA, Rowden AA, Attrill MJ (2001) Effect of patch size and in-patch location on the infaunal macroinvertebrate assemblages of Zostera marina seagrass beds. J Exp Mar Biol Ecol 259:133-154 Bowmaker JK (1995) The visual pigments of fish. Progress in Retinal and Eye Research 15:1-31 Bullard SJM, Hay ME (2002) Plankton tethering to assess spatial patterns of predation risk over a coral reef and seagrass bed. Mar Ecol Prog Ser 225:17-28 Cadenasso ML, Pickett STA, Weathers KC, Jones CG (2003) A framework for a theory of ecological boundries. BioScience 53:750-758 Cappo M, Harvey ES, Malcolm H, Speare P (2003) Potential of video techniques to monitor diversity, abundance and size of fish in studies of marine protected areas. In: Beumer JP, Grant A, Smith DC (eds) Aquatic Protected Areas - what works best and how do we know? World Congress on Aquatic Protected Areas. Australian Society of Fish Biology, Cairns, Australia,, p 455-464 Cappo M, Speare P, De'ath G (2004) Comparison of baited remote underwater video stations (BRUVS) and prawn (shrimp) trawls for assessments of fish biodiversity in inter-reefal areas of the Great Barrier Reef Marine Park. J Exp Mar Biol Ecol 302:123-152 Carruthers TJB, Walker DI (1997) Light climate and energy flow in the seagrass canopy of Amphibolis griffithii (J.M Black) den Hartog. Oecologia 109:335-341 Chittaro PM, Usseglio P, Sale PF (2005) Variation in fish density, assemblage composition and relative rates of predation among mangrove, seagrass and coral reef habitats. Environ Biol Fish 72:175-187 Connolly RM (1995) Effects of removal of seagrass canopy on assemblages of small, motile invertebrates. Mar Ecol Prog Ser 118:129-137

109 Connolly RM, Hindell JS (2006) Review of nekton patterns and ecological processes in seagrass landscapes. Est Coast Shelf Sci 68:433-444 Cronin JT (2009) Habitat edges, within-patch dispersion of hosts, and parasitoid oviposition behavior. Ecology 90:196-207 Dahlgren CP, Eggleston DB (2000) Ecological processes underlying ontogenetic habitat shifts in a coral reef fish. Ecology 81:2227-2240 Danilowicz BS, Sale PF (1999) Relative intensity of predation on the French grunt, Haemulon flavolineatum, during diurnal, dusk, and nocturnal periods on a coral reef. Mar Biol 133:337-343 Denny CM, Willis TJ, Babcock RC (2004) Rapid recolonisation of snapper Pragus auratus: Sparidae within an offshore island marine reserve after implementation of no-take status. Mar Ecol Prog Ser 272:183-190 Didham RK, Hammond PM, Lawton JH, Eggleton P, Stork NE (1998) Beetle species responses to tropical forest fragmentation. Ecol Monogr 68:295-323 Donovan TM, Jones PW, Annand EM, Thompsom FR (1997) Variation in local-scale edge effects: mechanisms and landscape context. Ecology 78:2064-2075 Duarte CM (2002) The future of seagrass meadows. Environmental Conservation 29:192- 206 Duarte CM, Chiscano CL (1999) Seagrass biomass and production: a reassessment. Aquatic Botany 65:159-174 Eckerman JE (1987) The role of hydrodynamics in recruitment, growth and survival of Argopecten irradians (L.) and Anomia simplex (D'Orbigny) within eelgrass meadows. J Exp Mar Biol Ecol 106:165-191 Edgar GJ, Shaw C (1995a) The production and trophic ecology of shallow-water fish assemblages in southern Australia 1. Species richness, size structure and production of fishes in , Victoria. J Exp Mar Biol Ecol 194:53-81 Edgar GJ, Shaw C (1995b) The production and trophic ecology of shallow-water fish assemblages in southern Australia II. Diets of fishes and trophic relationships between fishes and at Western Port, Victoria. J Exp Mar Biol Ecol 194:83- 106

110 Ewers RM, Didham RK (2006) Confounding factors in the detection of species responses to habitat fragmentation. Biol Rev 81:117-142 Ewers RM, Thorpe S, Didham RK (2007) Synergistic interactions between edge and area effects in a heavily fragmented landscape. Ecology 88:96-106 Fagan WE, Cantrell RS, Cosner C (1999) How habitat edges change species interactions. Am Nat 153:165-182 Fahrig L (2003) Effects of habitat fragmentation on biodiversity. Annu Rev Ecol Syst 34:487-515 Fell PE, Weissbach SP, Jones DA, Fallon MA, Zeppieri JA, Faison EK, Lennon KA, Newberry KJ, Reddington LK (1998) Does invasion of oligohaline tidal marshes by reed grass, Phragmites australis (Cav.) Trin. ex Steud., affect the availability of prey resources for the mummichog, Fundulus heteroclitus L.? J Exp Mar Biol Ecol 222:59-77 Ferrell DJ, Bell JD (1991) Differences among assemblages of fish associated with Zostera capricorni and bare sand over a large spatial scale. Mar Ecol Prog Ser 72:15-24 Fletcher RJ (2005) Multiple edge effects and their implications in fragmented landscapes. J Anim Ecol 74:342-352 Fletcher RJ, Ries L, Battin J, Chalfoun AD (2007) The role of habitat area and edge in fragmented landscapes: definitively distinct or inevitably intertwined? Canadian Journal of Zoology 85:1017-1030 Flynn AJ, Ritz DA (1999) Effect of habitat complexity and predatory style on the capture success of fish feeding on aggregated prey. J Mar Biol Ass U K 79:487-494 Fonseca MS, Fisher JS (1986) A comparison of canopy friction and sediment movement between four species of seagrass with reference to their ecology and restoration. Mar Ecol Prog Ser 29:15-22 Gibson RN, Robb L, Burrows MT, Ansell AD (1996) Tidal, diel and longer term changes in the distribution of fishes on a Scottish sandy beach. Mar Ecol Prog Ser 130:1-17 Gillanders BM (2007) Seagrass. In: Connell SD, Gillander BM (eds) Marine ecology. Oxford University Press, Melbourne, p 457-484

111 Gomon MF, Bray DJ, Kuiter RH (2008) Fishes of Australia's southern coast, Reed New Holland, Gorman AM, Gregory RS, Schneider DC (2009) Eelgrass patch size and proximity to the patch edge affect predation risk of recently settled age 0 cod (Gadus). J Exp Mar Biol Ecol 371:1-9 Haywood MDE, Manson FJ, Loneragan NR, Toscas PJ (2003) Investigation of artifacts from chronographic tethering experiments - interactions between tethers and predators. J Exp Mar Biol Ecol 290:271-292 Haywood MDE, Pendrey RC (1996) A new design for a submersible chronographic tethering device to record predation in different habitats. Mar Ecol Prog Ser 143:307-312 Heithaus MR, Hamilton IM, Wirsing AJ, Dill LM (2006) Validation of a randomization procedure to assess animal habitat preferences: microhabitat use of tiger sharks in a seagrass ecosystem. J Anim Ecol 75:666-676 Hettler WF (1989) Food habits of juveniles of spotted seatrout and gray snapper in western Florida Bay. Bull Mar Sci 44:155-162 Hinchey EK, Nicholson MC, Zajac RN, Irlandi EA (2008) Marine and coastal applications in landscape ecology. Landsc Ecol 23:1-5 Hindell JS (2006) Assessing the trophic link between seagrass habitats and piscivorous fishes. Mar Freshw Res 57:121-131 Hindell JS, Jenkins GP, Keough MJ (2000) Variability in abundances of fishes associated with seagrass habitats in relation to diets of predatory fishes. Mar Biol 136:725-737 Hindell JS, Jenkins GP, Keough MJ (2001) Spatial and temporal variability in the effects of fish predation on macrofauna in relation to habitat complexity and cage effects. Mar Ecol Prog Ser 224:231-250 Hindell JS, Jenkins GP, Keough MJ (2002) Variability in the numbers of post-settlement King George whiting (Sillaginidae: Sillaginodes punctata, Cuvier) in relation to predation, habitat complexity and artificial cage structure. J Exp Mar Biol Ecol 268:13-31

112 Hoese DF, Larson HK (2006) Description of two new species of Nesogobius (Pisces: Gobioidei: Gobiidae) from southern Australia. Memoir Vic Mus 63:7-13 Horinouchi M (2007a) Distribution patterns of benthic juvenile gobies in and around seagrass habitats: effectiveness of seagrass shelter against predators. Est Coast Shelf Sci 72:657-664 Horinouchi M (2007b) Review of within-patch scale structure complexity on seagrass fishes. J Exp Mar Biol Ecol 350:111-129 Hovel KA, Fonseca MS (2005) Influence of seagrass landscape structure on the juvenile blue crab habitat-survival function. Mar Ecol Prog Ser 300:179-191 Hovel KA, Fonseca MS, Myer DL, Kenworthy WJ, Whitfield PE (2002) Effects of seagrass landscape structure, structural complexity and hydrodynamic regime on macrofaunal densities in North Carolina seagrass beds. Mar Ecol Prog Ser 243:11- 24 Hovel KA, Lipcius RN (2002) Effects of seagrass habitat fragmentation on juvenile blue crab survival and abundance. J Exp Mar Biol Ecol 271:75-98 Howard RK, Koehn JD (1985) Population-dynamics and feeding ecology of pipefish (Syngnathidae) associated with eelgrass beds of Western-Port, Victoria. Aust J Mar Freshw Res 36:361-370 Hyndes GA, Kendrick AJ, MacArthur LD, Stewart E (2003) Differences in the species- and size-composition of fish assemblages in three distinct seagrass habitats with differing plant and meadow structure. Mar Biol 142:1195-1206 Irlandi EA (1997) Seagrass patch size and survivorship of an infaunal bivalve. Oikos 78:511-518 Irlandi EA, Orlando BA, Ambrose WG (1999) Influence of seagrass habitat patch size on growth and survival of juvenile bay scallops, Argopecten irradians concentricus (Say). J Exp Mar Biol Ecol 235:21-43 Jackson EL, Attrill MJ, Jones MB (2006a) Habitat characteristics and spatial arrangement affecting the diversity of fish and decapod assemblages of seagrass (Zostera marina) beds around the coast of Jersey (English Channel). Est Coast Shelf Sci 68:421-432

113 Jackson EL, Attrill MJ, Rowden AA, Jones MB (2006b) Seagrass complexity hierarchies: influence on fish groups around the coast of Jersey (English Channel). J Exp Mar Biol Ecol 330:38-54 Jackson EL, Rowden AA, Attrill MJ, Bossey SJ, Jones MB (2001) The importance of seagrass beds as a habitat for fishery species. Oceanogr Mar Biol Annu Rev 39:269- 303 Jackson EL, Rowden AA, Attrill MJ, Bossy SF, Jones MB (2002) Comparison of fish and mobile macroinvertebrates associated with seagrass and adjacent sand at St. Catherine Bay, Jersey (English Channel): Emphasis on commercial species. Bull Mar Sci 71:1333-1341 Jelbart JE, Ross PM, Connolly RM (2006) Edge effects and patch size in seagrass landscapes: an experimental test using fish. Mar Ecol Prog Ser 319:93-102 Jelbart JE, Ross PM, Connolly RM (2007) Fish assemblages in seagrass beds are influenced by the proximity of mangrove forests. Mar Biol 150:993-1002 Jenkins GP, Hamer PA (2001) Spatial variation in the use of seagrass and unvegetated habitats by post-settlement King George whiting (Percoidei: Sillaginidae) in relation to meiofaunal distribution and macrophyte structure. Mar Ecol Prog Ser 224:219- 229 Jenkins GP, May HMA, Wheatley MJ, Holloway MG (1997) Comparisons of fish assemblages associated with seagrass and adjacent unvegetated habitats of Port Phillip Bay and Corner Inlet, Victoria, Australia, with emphasis on commercial species. Est Coast Shelf Sci 44:569-588 Jenkins GP, Sutherland CR (1997) The influence of habitat structure on nearshore fish assemblages in a southern Australian embayment: colonisation and turnover rate of fishes associated with artificial macrophyte beds of varying physical structure. J Exp Mar Biol Ecol 218:103-125 Jenkins GP, Walker-Smith GK, Hamer PA (2002) Elements of habitat complexity that influence harpacticoid copepods associated with seagrass beds in a temperate bay. Oecologia 131:598-605

114 Jenkins GP, Wheatley MJ (1998) The influence of habitat structure on nearshore fish assemblages in a southern Australian embayment: Comparison of shallow seagrass, reef-algal and unvegetated sand habitats, with emphasis on their importance to recruitment. J Exp Mar Biol Ecol 221:147-172 Johnson MW, Heck KL (2006) Effects of habitat fragmentation per se on decapods and fishes inhabiting seagrass meadows in the northern Gulf of Mexico. Mar Ecol Prog Ser 306:233-246 Kendrick AJ, Hyndes GA (2005) Variations in the dietary compositions of morphologically diverse syngnathid fishes. Environ Biol Fish 72:415-427 Kremsater LL, Bunnell FL (1992) Testing Responses to Forest Edges - the Example of Black-Tailed Deer. Can J Zool-Rev Can Zool 70:2426-2435 Laurance WF, Ferreira LV, Rankin-De Merona JM, Laurance SG (1998) Rain forest fragmentation and the dynamics of Amazonian tree communities. Ecology 79:2032- 2040 Laurel BJ, Gregory RS, Brown JA (2003) Predator distribution and habitat patch area determine predation rates on Age-0 juvenile cod Gadus spp. Mar Ecol Prog Ser 251:245-254 Laurel BJ, Stoner AW, Ryer CH, Hurst TP, Abookire AA (2007) Comparative habitat associations in juvenile Pacific cod and other gadids using seines, baited cameras and laboratory techniques. J Exp Mar Biol Ecol 351:42-55 Lecomte N, Careau V, Gauthier G, Giroux J (2008) Predator behaviour and predation risk in the heterogeneous Arctic environment. J Anim Ecol 77:439-447 Leopold A (1933) Game management, Charles Scribner's Sons, New York Macreadie P, Hindell J, Keough M, Jenkins G, Connolly R (In press) Resource distribution influences positive edge effects in a seagrass fish. Ecology Macreadie PI, Hindell JS, Jenkins GP, Connolly RM, Keough MJ (2009) Fish responses to experimental fragmentation of seagrass habitat. Conservat Biol 23:644-652 Malcolm JR (1994) Edge effects in central Amazonian forest fragments. Ecology 75:2438- 2445

115 Meyer CL, Sisk TD, Covington WW (2001) Microclimatic changes induced by ecological restoration of Ponderosa Pine Forests in Northern Arizona. Restor Ecol 9:443-452 Mills DJ, Jonhnson CR, Gardner C (2008) Bias in lobster tethering experiments conducted for selecting low-predation release site. Mar Ecol Prog Ser 364:1-13 Mosk V, Thomas N, Hart NS, Partridge JC, Beazley LD, Shand J (2007) Spectral sensitivities of the seahorses Hippocampus subelongatus and Hippocampus barbouri and the pipefish Stigmatopora argus. Visual Neuroscience 24:345-354 Murcia C (1995) Edge effects in fragmented forests: implications for conservation. Trends Ecol Evol 10:58-62 Nagelkerken I, Dorenbosch M, Verberk W, de la Moriniere EC, van der Velde G (2000) Day-night shifts of fishes between shallow-water biotopes of a Caribbean bay, with emphasis on the nocturnal feeding of Haemulidae and Lutjanidae. Mar Ecol Prog Ser 194:55-64 Nel R, McLachlan A, Winter DPE (2001) The effect of grain size on the burrowing of two Donax species. J Exp Mar Biol Ecol 265:219-238 Orth RJ, Carruthers TJB, Dennison WC, Duarte CM, Fourqurean JW, Heck KL, Hughes AR, Kendrick GA, Kenworthy WJ, Olyarnik S, Short FT, Waycott M, Williams SL (2006a) A global crisis for seagrass ecosystems. BioScience 56:987-996 Orth RJ, Heck KLJ, von Montfrans J (1984) Faunal communities in seagrass beds: a review of the influence of plant structure and prey characteristics on predator-prey relationships. Estuaries 7:339-350 Orth RJ, Kendrick GA, Marion SR (2006b) Predation on Posidonia australis seeds in seagrass habitats of Rottnest Island, Western Australia: patterns and predators. Mar Ecol Prog Ser 313:105-114 Parker TH, Stansberry BM, Becker CD, Gipson PS (2005) Edge and area effects on the occurence of migrant forest songbirds. Conservat Biol 19:1157-1167 Paton PWC (1994) The effect of edge on avian nest success - how strong is the evidence. Conservat Biol 8:17-26

116 Peterson BJ, Thompson KR, Cowan JH, Heck KL (2001) Comparison of predation pressure in temperate and subtropical seagrass habitats based on chronographic tethering. Mar Ecol Prog Ser 224:77-85 Peterson CH, Black R (1994) An experimentalists challenge - when artifacts of intervention interact with treatments. Mar Ecol Prog Ser 111:289-297 Peterson CH, Luettich RA, Micheli F, Skilleter GA (2004) Attenuation of water flow inside seagrass canopies of differing structure. Mar Ecol Prog Ser 268:81-92 Quinn GP, Keough MJ (2002) Experimental design and data analysis for biologists, Vol. Cambridge University Press, Cambridge Ries L, Fletcher RJ, Battin J, Sisk T, D. (2004) Ecological responses to habitat edges: Mechanisms, models and variability explained. Annu Rev Ecol Syst 35:491-522 Ries L, Sisk T, D. (2004) A predictive model of edge effects. Ecology 85:2917-2962 Robbins BD, Bell SS (1994) Seagrass landscapes - a terrestrial approach to the marine subtidal environment. Trends Ecol Evol 9:301-304 Robertson AI (1980) The structure and organization of an eelgrass fish fauna. Oecologia 47:76-82 Robertson AI (1982) Population-dynamics and feeding ecology of juvenile Australian salmon (Arripis-Trutta) in Western-Port, Victoria. Aust J Mar Freshw Res 33:369- 375 Rypel AL, Layman CA, Arrington DA (2007) Water depth modifies relative predation risk for a motile fish taxon in Bahamian tidal creeks. Estuaries and Coasts 30:1-8 Sanchez-Jerez P, Barbera-Cebrian C, Ramos-Espla A (1999a) Daily vertical migrations in the epifauna associated with Posidonia oceanica meadows. J Mar Biol Ass U K 79:971-977 Sanchez-Jerez P, Cebrian CB, Espla AAR (1999b) Comparison of the epifauna spatial distribution in Posidonia oceanica, Cymodocea nodosa and unvegetated bottoms: importance of meadow edges. Acta Oecol Int J Ecol 20:391-405 Scoffin TP (1970) The trapping and binding of subtidal carbonate sediments by marine vegetation in Bimini Lagoon, Bahamas. Journal of Sediment Petrology 40:249-273

117 Sheaves M (2001) Are there really few piscivorous fishes in shallow estuarine habitats? Mar Ecol Prog Ser 222:279-290 Smith TM, Hindell JS (2005) Assessing effects of diel period, gear selectivity and predation on patterns of microhabitat use by fish in a mangrove dominated system in SE Australia. Mar Ecol Prog Ser 294:257-270 Smith TM, Hindell JS, Jenkins GP, Connolly RM (2008) Edge effects on fish associated with seagrass and sand patches. Mar Ecol Prog Ser 359:203-213 Smith TM, Hindell JS, Jenkins GP, Connolly RM (2010) Seagrass patch size affects fish responses to edges. J Anim Ecol 79:275-281 Sogard SM, Able KW (1993) A comparison of eelgrass, sea lettuce, macroalgae and marsh creeks as habitats for epibenthic fishes and decapods. Est Coast Shelf Sci 33:501- 519 Steffe AS, Westoby M, Bell JD (1989) Habitat selection and diet in two species of pipefish from seagrass: sex differences. Mar Ecol Prog Ser 55:23-30 Stobart B, Garcia-Charton JA, Espejo C, Rochel E, Goni R, Renones O, Herrero A, Crec'hriou R, Polti S, Marcos C, Planes S, Perez-Ruzafa A (2007) A baited underwater video technique to assess shallow-water Mediterranean fish assemblage: methodological evaluation. J Exp Mar Biol Ecol 345:158-174 Strayer DL, Power ME, Fagan WE, Pickett STA, Belnap J (2003) A classification of ecological boundaries. BioScience 53:723-729 Tanner JE (2003) Patch shape and orientation influences on seagrass epifauna are mediated by dispersal abilities. Oikos 100:517-524 Tanner JE (2005) Edge effects on fauna in fragmented seagrass meadows. Austral Ecology 30:210-218 Tanner JE (2006) Landscape ecology of interactions between seagrass and mobile epifauna: The matrix matters. Est Coast Shelf Sci 68:404-412 Turner MG (1989) Landscape ecology: the effect of pattern on process. Annu Rev Ecol Syst 20:171-197 Uhrin AV, Holmquist JG (2003) Effects of propeller scarring on macrofaunal use of the seagrass Thalassia testudinum. Mar Ecol Prog Ser 250:61-70

118 Walters K, Bell SS (1986) Diel patterns of active vertical migration in seagrass meiofauna. Mar Ecol Prog Ser 34:95-103 Warry FY, Hindell JS, Macreadie PI, Jenkins GP, Connolly RM (2009) Integrating edge effects into studies of habitat fragmentation: a test using meiofauna in seagrass. Oecologia 159:883-892 Watson DL, Harvey ES, Anderson MJ, Kendrick AJ (2005) A comparison of temperate reef fish assemblages recorded by three underwater stereo-video techniques. Mar Biol 148:415-425 Widder EA, Robison BH, Reisenbichler KR, Haddock SHD (2005) Using red light for in situ observations of deep-sea fishes. Deep-Sea Research 52:2077-2085 Willis TJ, Babcock RC (2000) A baited underwater video system for the determination of relative density of carnivorous reef fish. Mar Freshw Res 51:755-763 Willis TJ, Millar RB, Babcock RC (2000) Detection of spatial variability in relative density of fishes: comparison of visual census, angling, and baited underwater video. Mar Ecol Prog Ser 198:249-260 Young A, Mitchell N (1994) Microclimate and vegetation edge effects in a fragmented podocarp-broadleaf forest in New Zealand. Biol Conserv 67:63-72

119

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Smith, Timothy Malcolm

Title: The importance of edge effects in determining fish distributions in patchy seagrass habitats

Date: 2009

Citation: Smith, T. M. (2009). The importance of edge effects in determining fish distributions in patchy seagrass habitats. PhD thesis, Faculty of Science, Zoology, Victorian Marine Science Consortium, The University of Melbourne.

Publication Status: Published

Persistent Link: http://hdl.handle.net/11343/35306

File Description: The importance of edge effects in determining fish distributions in patchy seagrass habitats

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.