Patterns in the Distribution and Abundance of Reef Fishes in South Eastern Australia

Home , Acanthopagrus australis, Achoerodus gouldii, Achoerodus viridis, Atypichthys, Atypichthys strigatus, Caesioperca

Madhavi A. Colton

Submitted in total fulfilment of the requirements of the degree of Doctor of Philosophy

March 2011

Department of Zoology The University of Melbourne Parkville, Victoria 3010 Australia

ABSTRACT

This research investigated patterns in the distribution and abundance of nearshore fishes of south-eastern Australia. I used two methods to survey fishes, underwater visual census (UVC) and baited remote underwater video (BRUV). A comparison between these methods revealed that BRUV recorded higher relative abundance of mobile predators, while UVC observed higher relative abundance of herbivores, territorial species, and small site-attached species. These results suggest that studies surveying diversity would do best to employ multiple methods. In cases where funds are limited, UVC may provide a more complete estimate of diversity than BRUV as UVC recorded higher diversity, species richness and more individuals.

Combining measures of abundance with habitat data, I investigated fish-habitat associations, specifically exploring how altering spatial grain influenced the strength of correlations between fish and habitat. Species of different sizes responded to habitat measured over different scales, with large-bodied species only displaying strong correlations with habitat when it was measured over large scales. These results suggest that research quantifying fish-habitat associations needs to take spatial grain into account. In addition, many species may respond to changes in habitat at scales larger than are typically investigated. Understanding not only how species interact with their environment but also the scale at which these associations occur is essential for management and conservation.

I investigated biogeographic patterns in the distribution of fishes in Victoria using abundance measured by BRUV and UVC. The BRUV data displayed a cline in change across the state in which dissimilarities between locations were linearly related to distance. In contrast, data collected using UVC indicated the presence of a large faunal break in the vicinity of Ninety Mile Beach, and a second break between Cape Conran and Cape Howe, suggesting that contemporary habitat discontinuity, flow and/or temperature may be important factors structuring communities in this region.

At a still larger scale, I explored relationships at upper and lower bounds between body size, geographic range size and abundance using data collected from Australia and New Zealand. At maxima, the relationship between body size and abundance was

i negative but steeper than expected, possibly driven by diver-averse behaviour of large species. At minima, body size and geographic range size were positively related, implying that body size determines the minimum area that a species must occupy. In contrast, at the upper bound this relationship was negative for non-perciform fishes, a K-selected group whose geographic range size could be constrained by their limited dispersal capacity. Distribution-abundance relationships deviated from predictions, with a negative relationship at the upper bound for Perciformes, which could be driven by the high dispersal potential of widespread species that results in diffuse low- density populations. From these results, I concluded, first, that fishes appear to differ from terrestrial taxa, which may be attributed in part to large-bodied fishes’ limited capacity for dispersal. Second, the approach of applying regressions to maxima and minima uncovered relationships that would have been obscured had they been investigated at the mean, highlighting the importance of exploring limits in macroecological relationships.

DECLARATION

This is to certify that

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

7 March, 2011

Madhavi A. Colton Date

iii

PREFACE

Chapter Four is a meta-analysis of data collected by myself, Dr. Sean Connell of the University of Adelaide, and Mr. Alejandro Pérez-Matus of Victoria University in Wellington, New Zealand. I conceived the idea for the chapter, analysed and wrote up the data. Dr. Connell and Mr. Pérez-Matus supplied measures of fish density that were made in the field.

ACKNOWLEDGEMENTS

Thanks are due first and foremost to my supervisor Dr. Stephen Swearer whose advice and support were invaluable. This work would not have been feasible without the field assistance of Mr. Dean Chamberlain, Mr. John Ford, Dr. Christian Jung, Mr. Matthew LeFeuvre and Mr. Malcolm Lindsay. Mr. Michael Sams was immensely helpful in identifying the invertebrates, algae and seagrasses in habitat photos. The Parks Victoria staff at the Mallacoota Office went above and beyond in their support of our research efforts at Gabo Island. Thanks also to the Parks Victoria crew at Wilsons Promontory. This research was funded by Natural Heritage Trust and Parks Victoria, and I was supported by an Australian Postgraduate Award and a Helen McPherson-Smith Scholarship. I was also supported by my husband Zack Kushner who showed incredible patience with me through this process.

TABLE OF CONTENTS General Introduction…………………………………………………………… 1 Literature Cited…………………………………………………………….. 3

Chapter One: A comparison of two survey methods: differences between underwater visual census and baited remote underwater video……………..... 6 Abstract……………………………………………………………...... 6 Introduction………………………………………………………….……… 7 Materials & Methods……………………………………………………….. 12 Results………………………………………………………………………. 25 Discussion…………………………………………………………...... 30 Conclusion …………………………………………………………………. 40 Literature Cited……………………………………………………………... 40

Chapter Two: Body size and spatial grain influence fish-habitat associations in temperate marine fishes ……………………………………………………… 44 Abstract……………………………………………………………...... 44 Introduction………………………………………………………….……… 44 Materials & Methods………………………………………………………... 48 Results………………………………………………………………………. 54 Discussion…………………………………………………………...... 60 Literature Cited……………………………………………………………... 68

Chapter Three: Locating faunal breaks in the nearshore fish assemblage of Victoria, Australia…...………………………………………………………….. 71 Abstract……………………………………………………………...... 71 Introduction………………………………………………………….……… 71 Materials & Methods……………………………………………………….. 76 Results………………………………………………………………………. 82 Discussion…………………………………………………………...... 87 Literature Cited……………………………………………………………... 96

Chapter Four: Relationships between Geographic Range Size, Body Size and Abundance in Temperate Marine Fishes…………………..…………………. 98 Abstract……………………………………………………………...... 98 Introduction………………………………………………………….……… 99 Materials & Methods……………………………………………………….. 102 Results………………………………………………………………………. 110 Discussion…………………………………………………………...... 112 Conclusion …………………………………………………………………. 124 Literature Cited……………………………………………………………... 125

General Conclusion……………………………………………………………. 129 Literature Cited…………………………………………………………….. 130

GENERAL INTRODUCTION

Ecology is fundamentally the study of the distribution and abundance of species, and these data are essential to any management or conservation plan. Numerous factors can influence where a species occurs and the densities its populations attain. Results from ecological studies investigating the mechanisms driving abundance and distribution are “inextricably scale-dependent” (Sale 1998). At the scale of 1000s km, for example, the distribution of fishes may be best explained by currents and their associated temperatures (e.g. Figueira & Booth 2010), while at the scale of a single reef, a species’ distribution may be strongly associated with structural complexity (e.g. Leum & Choat 1980, Wiens 1989). To gain a complete understanding of abundance and distribution requires investigating these parameters at multiple scales (Eagle et al. 2001, Meyer & Thuiller 2006, Thogmartin & Knutson 2007, Grober- Dunsmore et al. 2008). In particular, the terrestrial literature indicates that consideration of the scale at which an organism relates to its environment is important (Meyer & Thuiller 2006, Murakami et al. 2008), which is ultimately determined by its body size (Holling 1992). There is considerable evidence in the terrestrial literature, and some in the marine literature, to suggest that body size places an ultimate constraint on density (Mohr 1940, Damuth 1981, Marquet et al. 1990, Cotgreave 1993, McGill 2008) and is positively correlated with geographic range size (Brown 1995). In addition, at least in terrestrial systems, range size and abundance are also frequently related (Brown 1984, Borregaard & Rahbek 2010).

More than just body size, however, determines how an organism relates to its environment. Behavioural attributes and life history interact to determine where a species can occur and it population growth. The study of distribution, or range size, is ultimately the study of dispersal and colonization (Myers 1997). Dispersal capacity and colonization success are a function of interactions between species-specific traits, e.g. reproductive mode and body size (e.g. Reaka 1980), and the environment (Marshall et al. 2010). Whether dispersal is a good predictor of range size in the marine environment is a matter of some debate (Jones et al. 2002, Hawkins et al. 2007, Pelc et al. 2009, Weersing & Toonen 2009). However, in terrestrial systems, there does appear to be a link between abundance and connectivity as populations tend to decline in fragmented habitats (Gonzalez et al. 1998). Colonization can only

- 1 - occur in areas where a species’ basic resource requirements are met and sites with abundant resources may support larger populations. Multi-species interactions can be very influential in determining a species’ abundance and distribution (Wethey 2002, White 2007, Sexton et al. 2009), by facilitating or inhibiting range occupation and population growth. Interactions with Homo sapiens most frequently negatively impact other species’ population densities and ranges, either through habitat fragmentation or resource dominance. A less frequently considered way in which humans impact species’ abundances and distributions is through our observations. The methods we choose to measure these data will influence our perception of these parameters.

Rocky reefs support a wide variety of flora and fauna, including many species of bony and cartilaginous fishes. Much of our understanding of the ecology of subtidal temperate environments comes from the Northern Hemisphere, where systems are often upwelling-dominated (e.g., the California Current) or impacted by a long history of fishing (e.g., the North Atlantic). The state of Victoria in south-east Australia has a dynamic coastline that functions as a convergence zone for several distinct marine bioregions (Whitely 1932, Hough & Mahon 1994, Lyne et al. 1996). The East Australian Current carries warmer water from the tropics south to the Tasman Sea, with eddies moving as far south as Tasmania (Gibbs 1991, Ridgway & Dunn 2003, Middleton & Cirano 2005). The western part of the state is influenced by the South Australian Current, which flows east from Western Australia (Ridgway & Condie 2004). In the centre lies Bass Strait, which is a fairly stagnant body of water due to the actions of currents and tides (Fandry 1983, Baines et al. 1991, Gibbs 1991, Evans & Middleton 1998).

Victoria’s dynamic oceanography gives this region a high potential for the emergence of interesting patterns in abundance and distribution. Research in this region has primarily focused on the biogeography and phylogeography of macroinvertebrates (e.g. O'Hara & Poore 2000, O'Hara 2001, O'Loughlin et al. 2003, Waters & Roy 2003, Waters et al. 2004, Dawson 2005, Waters et al. 2005, Hidas et al. 2007, Waters 2008). However, there are a few studies that suggest that the fishes of this region also exhibit biogeographic structure. For example, the life history of the shark Heterodontus portusjacksoni differs between eastern and western Victoria (Tovar- Ávila et al. 2007), and there are several species pairs with east-west distributions that - 2 - may have arisen as a result of vicariance during glacial periods, e.g. Paraplesiops spp. (Hutchins 1987). Finally, there are many species whose ranges terminate in eastern Victoria (Kuiter 2000, Edgar 2005, Gomon et al. 2008), though it is unclear whether these terminations occur primarily at Wilsons Promontory or further east.

In this research, I investigated patterns in the abundance and distribution of Victoria’s rocky reef ichthyofauna. I was specifically interested in understanding the effect that species-specific attributes, such as behaviour and body size, have upon abundance and distribution. I began by exploring how two survey methods, underwater visual census (UVC) and baited remote underwater video (BRUV), compared in the types and numbers of species they record, and how species’ behaviour and size influenced these methods (Chapter One). In Chapter Two, I discuss results from a study that quantified associations between fish and habitat, and that specifically examined how body size and spatial grain interact to determine species’ habitat associations. From this research it became apparent that there were larger-scale factors influencing species’ distributions in Victoria. In Chapter Three, I discuss the biogeography of the fishes of south-eastern Australian with regard to contemporary and historical barriers, currents and temperature, identifying the site of two biogeographical disjunctions in eastern Victoria. The final chapter, Chapter Four, investigates macroecological relationships between body size, abundance and distribution for fishes of southern Australia and New Zealand, and is the first exploration of relationships between these variables for temperate marine fishes.

LITERATURE CITED Baines PG, Hubbert G, Power S (1991) Fluid transport through Bass Strait. Continental Shelf Research 11:269-293 Borregaard MK, Rahbek C (2010) Causality of the relationship between geographic distributio and species abundance. Quarterly Review of Biology 85:3-25 Brown JH (1984) On the relationship between abundance and distribution of species. The American Naturalist 124:255-279 Brown JH (1995) Macroecology, University of Chicago Press, Chicago Cotgreave P (1993) The relationship between body size and population abundance in animals. Trends in Ecology and Evolution 8:244-248 Damuth J (1981) Population density and body size in mammals. Nature 290:699-700 Dawson MN (2005) Incipient speciation of Catostylus mosaicus (Scyphozoa, Rhizostomeae, Catostylidae), comparative phylogeography and biogeography in south-east Australia. Journal of Biogeography 32:515-533

- 3 -

Eagle JV, Jones GP, McCormick MI (2001) A multi-scale study of the relationships between habitat use and the distribution and abundance patterns of three coral reef angelfishes (Pomacanthidae). Marine Ecology Progress Series 214:253-265 Edgar GJ (2005) Australian Marine Life: The Plants and Animals of Temperate Waters, Reed New Holland Publishers Pty Ltd, Sydney, Australia Evans SR, Middleton JF (1998) A regional model of shelf circulation near Bass Strait: A new upwelling mechanism. Journal of Physical Oceanography 28:1439-1457 Fandry CB (1983) Model for the 3-dimensional structure of wind-driven and tidal circulation in Bass Strait. Australian Journal of Marine and Freshwater Research 34:121-141 Figueira WF, Booth DJ (2010) Increasing ocean temperatures allow tropical fishes to survive overwinter in temperate waters. Global Change Biology 16:506-516 Gibbs CF (1991) Oceanography of Bass Strait - implication for the food supply of Little Penguins Eudyptula minor. The Emu 91:395-401 Gomon MF, Bray D, Kuiter R (2008) Fishes of Australia's Southern Coast, Reed New Holland, Sydney Gonzalez A, Lawton JH, Gilbert FS, Blackburn TM, Evans-Freke I (1998) Metapopulation dynamics, abundance, and distribution in a microecosystem. Science 281:2045-2047 Grober-Dunsmore R, Frazer TK, Beets JP, Lindberg WJ, Zwick P, Funicelli NA (2008) Influence of landscape structure, on reef fish assemblages. Landscape Ecolog 23:37-53 Hawkins BA, Diniz-Filho JAF, Jaramillo CA, Soeller SA (2007) Climate, Niche Conservatism, and the Global Bird Diversity Gradient. The American Naturalist 170:S16- S27 Hidas EZ, Costa TL, Ayre DJ, Minchinton TE (2007) Is the species composition of rocky intertidal invertebrates across a biogeographic barrier in south-eastern Australia related to their potential for dispersal? Marine and Freshwater Research 58:835-842 Holling CS (1992) Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62:447-502 Hough D, Mahon G (1994) Biophysical classification of Victoria's marine waters. In: Muldoon J (ed) Towards a Marine Regionalisation for Australia. Great Barrier Reef Marine Park Authority, Sydney Hutchins JB (1987) Description of a new plesiopid fish from south-western Australia with a discussion of the zoogeography of Paraplesiops. Records of the Western Australian Museum 13:231-240 Jones GP, Caley MJ, Munday PL (2002) Rarity in coral reef fish communities. In: Sale PF (ed) Coral Reef Fishes: Dynamics and Diversity in a Complex Ecosystem. Academic Press, San Diego, USA, p 81-101 Kuiter RH (2000) Coastal Fishes of South-eastern Australia, Gary Allen Pty Ltd, Sydney Leum LL, Choat JH (1980) Density and distribution patterns of the temperate marine fish Cheilodactylus spectabilis (Cheilodactylidae) in a reef environment. Marine Biology 57:327-337 Lyne V, Last PR, Gomon MF, Scott R, Long S, Phillips A, McArdle B, Peters D, Pigot S, Kailola P (1996) Interim Marine Bioregionalisation for Australia: Towards a National System of Marine Protected Areas. CSIRO Division of Fisheries and Division of Oceanography, p 73 Marquet PA, Navarrete SA, Castilla JC (1990) Scaling population density to body size in rocky interridal communities. Science 250:1125-1127 Marshall DJ, Monro K, Bode M, Keough MJ, Swearer S (2010) Phenotype-environment mismatches reduce connectivity in the sea. Ecology Letters 13:128-140 McGill BJ (2008) Exploring predictions of abundance from body mass using hierarchical comparative approaches. American Naturalist 172:88-101 Meyer CB, Thuiller W (2006) Accuracy of resource selection functions across spatial scales. Diversity and Distribution 12:288-297 Middleton JF, Cirano M (2005) Wintertime circulation off southeast Australia: Strong forcing by the East Australian Current. Journal of Geophysical Research - Oceans 110:12

- 4 -

Mohr CO (1940) Comparative Populations of Game, Fur and Other Mammals. American Midland Naturalist 24:581-584 Murakami M, Hirao T, Iwamoto J, Oguma H (2008) Effects of windthrow disturbance on a forest bird community depend on spatial scale. Basic and Applied Ecology 9:762-770 Myers AA (1997) Biogeographic barriers and the development of marine biodiversity. Estuarine and Coastal Shelf Science 44:241-248 O'Hara TD (2001) Patterns of Diversity for Subtidal Reef Assemblages of Victoria, Australia. PhD, The University of Melbourne O'Hara TD, Poore GCB (2000) Patterns of distribution for southern Australian marine echinoderms and decapods. Journal of Biogeography 27:1321-1335 O'Loughlin PM, Waters JM, Roy MS (2003) A molecular and morphological review of the asterinid, Patiriella gunnii (Gray) (Echinodermata: Asteroidea). Memoirs of Museum Victoria 60:181-195 Pelc RA, Warner RR, Gaines SD (2009) Geographical patterns of genetic structure in marine species with contrasting life histories. Journal of Biogeography 36:1881-1890 Reaka ML (1980) Geographic range, life history patterns, and body size in a guild of coral- dwelling Mantis shrimps. Evolution 34:1019-1030 Ridgway KR, Condie SA (2004) The 5500-km-long boundary flow off western and southern Australia. Journal of Geophysical Research 109:18 pp.-18 pp. Ridgway KR, Dunn JR (2003) Mesoscale structure of the mean East Australian Current System and its relationship with topography. Progress in Oceanography 56:189-222 Sale PF (1998) Appropriate spatial scales for studies of reef-fish ecology. Australian Journal of Ecology 23:202-208 Sexton JP, McIntyre PJ, Angert AL, Rice KJ (2009) Evolution and ecology of species range limits. Annual Review of Ecology Evolution and Systematics 40:415-436 Thogmartin WE, Knutson MG (2007) Scaling local species-habitat relations to the larger landscape with a hierarchical spatial count model. Landscape Ecology 22:61-75 Tovar-Ávila J, Walker TI, Day RW (2007) Reproduction of Heterodontus portusjacksoni in Victoria, Australia: evidence of two populations and reproductive parameters. Marine and Freshwater Research 58:956-965 Waters JM (2008) Marine biogeographical disjunction in temperate Australia: historical landbridge, contemporary currents, or both? Diversity and Distribution 14:692-700 Waters JM, King TM, O'Loughlin PM, Spencer HG (2005) Phylogeographical disjunction in abundant high-dispersal littoral gastropods. Molecular Ecology 14:2789-2802 Waters JM, O'Loughlin PM, Roy MS (2004) Cladogenesis in a starfish species complex from southern Australia: evidence for vicariant speciation? Molecular Phylogenetics and Evolution 32:236-245 Waters JM, Roy MS (2003) Marine biogeography of southern Australia: phylogeographical structure in a temperate sea-star. Journal of Biogeography 30:1787-1796 Weersing K, Toonen RJ (2009) Population genetics, larval dispersal, and connectivity in marine systems. Marine Ecology-Progress Series 393:1-12 Wethey DS (2002) Biogeography, competition, and microclimate: The barnacle Chthamalus fragilis in New England. Integrative and Comparative Biology 42:872-880 White JW (2007) Spatially correlated recruitment of a marine predator and its prey shapes the large-scale pattern of density-dependent prey mortality. Ecology Letters 10:1054-1065 Whitely G (1932) Marine zoogeographical regions of Australia. The Australian Naturalist 8:166-167 Wiens JA (1989) Spatial scaling in ecology. Functional Ecology 3:385-397

- 5 -

CHAPTER ONE

A comparison of two survey methods: differences between underwater visual census and baited remote underwater video

Published in Marine Ecology-Progress Series Vol. 400: 19-36, 2010

ABSTRACT Essential to any model, conservation or management plan are measures of the distribution and abundance of a species. Countless methods for estimating these parameters exist, making it essential to assess the limitations and biases associated with a particular sampling protocol. Here, we compare between 2 methods commonly used to survey nearshore fish assemblages. Although most commonly employed, underwater visual census (UVC) may yield biased estimates of abundance depending on the strength of a fish’s behavioural response (i.e. avoidance, attraction) to the presence of divers. Baited remote underwater video (BRUV) techniques have shown promise in overcoming some of the limitations of UVC, but are unable to provide an absolute measure of density in turbulent environments. We compare the abilities of these 2 methods to survey the nearshore rocky reef ichthyofauna of Southeast Australia. We found that relative to BRUV, UVC recorded more individuals (in terms of all species, herbivores, cryptic species, and most territorial species), higher richness at both the species and family level, and higher biodiversity as measured using the Shannon Index. These findings remain even when the data were adjusted for differences in sampling effort. In contrast, BRUV recorded proportionally more mobile predators, and a more taxonomically distinct population, though only when taxonomic evenness was not taken into account. Twenty species were unique to UVC and 17 species unique to BRUV. Considering this, studies aimed at cataloguing diversity should apply multiple methods. However, when logistical or financial constraints limit biodiversity studies to only 1 method, UVC will likely provide a more complete estimate of temperate reef fish diversity than BRUV.

KEY WORDS: Subtidal fish assemblages; Rocky reefs; Diversity; Herbivores; Territoriality; Mobile predators; Sightability index; Taxonomic distinctness; Victoria; Australia - 6 -

INTRODUCTION Estimates of the abundance and distribution of a species are essential to any ecological model, though measures of these parameters will often be influenced by the method chosen to obtain these data. While the sources of bias within a research protocol may vary, the presence of bias is inevitable (MacNeil et al. 2008a), leaving it to researchers to identify sources of bias within their own data. In the marine environment, accurately assessing population abundance is very challenging. Researchers are limited in the amount of time they can spend monitoring a population and frequently find themselves in conditions that make observations difficult. Considering this, it is unsurprising that numerous techniques have been developed to sample subtidal marine communities, making it essential to compare between methods in order to identify their relative biases.

Bias in surveys can be caused by factors that are intrinsic to the species being observed, as well as by the survey methodology itself. Although most methodologies assume that detectability, or the probability of observing a species, is equal for all species, in reality, the fauna of any region is comprised of types of species that are variably detectable (Boulinier et al. 1998, MacNeil et al. 2008a, MacNeil et al. 2008b). MacNeil et al. (2008a) found that the greatest source of heterogeneity in detectability was caused by species characteristics, i.e. physical traits, behaviour, and life history. To compare between survey methods and identify biases requires classifying the types of species that are differentially detected by each method. In this research, we explore how intrinsic factors influence the detection capacity of 2 methods commonly used to survey subtidal fish assemblages: underwater visual census (UVC) and baited remote underwater video (BRUV).

BRUV has been used to survey species for at least 30 yr (Miller 1975), though its application has increased markedly in recent years, particularly in Australia (e.g. Cappo et al. 2007). BRUV is particularly effective at recording a diverse assemblage of species (Willis & Babcock 2000, Cappo et al. 2004, Watson et al. 2005). However, BRUV units deployed in turbulent environments are unable to provide measures of absolute density because such measures necessitate that the area surveyed be quantified. This requires detailed information about the dynamics of the bait odour plume, and the sensory capacity, swimming speed, and behaviour of individual - 7 -

Figure 1: The number of surveys, n, conducted using each method at locations along the coast of Victoria in south eastern Australia. An inset of one location, Wilsons Promontory, is provided to illustrate the positioning of replicate BRUV deployments relative to the paired UVC, with black squares indicating BRUV replicates, white triangles indicating the start of UVC transects, and grey triangles the end of UVC transects. species (Priede & Merrett 1996, Yau et al. 2001, Farnsworth et al. 2007, Heagney et al. 2007), data which are often unavailable for many fishes. The result is that BRUV

- 8 - typically can only provide a measure of relative abundance, particularly in turbulent areas.

In comparison, UVC has a long history of application in surveys of fish assemblages (Brock 1954). The problems associated with the use of UVC to estimate species’ densities have been well documented and can be caused by a variety of factors including but not limited to a species’ cryptic colouration or secretive behaviour (Sale & Douglas 1981, Kulbicki 1998), a diver’s ability to identify and accurately count individuals (Harvey et al. 2004), and a fish’s behavioural response to divers (Kulbicki 1998, Watson et al. 2007). The result of these biases is that UVC usually underestimates fish density (Sale & Sharp 1983, Edgar et al. 2004).

Few studies have compared UVC to BRUV, and most of these have found BRUV to be superior to UVC at measuring diversity (Willis & Babcock 2000, Willis et al. 2000, Watson et al. 2005), though at least one study has found the opposite to be true (Stobart et al. 2007). To the best of our knowledge, no studies have examined how these 2 methods compare in measures of taxonomic diversity. If BRUV is better at recording mobile predators (Willis & Babcock 2000, Cappo et al. 2004, Watson et al. 2005, Watson & Harvey 2007), including many elasmobranchs, then it may survey a more diverse assemblage. If, however, UVC is equally good at surveying these species (Friedlander & DeMartini 2002, Castro & Rosa 2005, Powter & Gladstone 2008) then there may be little difference between the methods in measures of taxonomic distinctness. In this research, we were interested not only in how these 2 methods differ in various measures of diversity and richness, but also in how factors that are intrinsic to a species may influence how well a method detects a species. Specifically we were interested in the effects of conspicuousness, crypsis, mobility, territoriality, trophic position and schooling behaviour on detectability using BRUV and UVC.

Conspicuous species are perhaps the easiest to survey, with numerous methods developed to measure their densities. However, being conspicuous alone is not enough to guarantee observation. For example, MacNeil et al. (2008a) found that scarids had a relatively low probability of detection despite their physical conspicuousness, and attributed this incongruity to the species’ highly mobile nature - 9 - and diver-averse behaviour. The behaviour of many fishes changes in the presence of divers (Kulbicki 1998, Watson & Harvey 2007), which may explain why some researchers have found BRUV to be especially effective at observing predatory and mobile species, such as elasmobranchs or species that are targeted by fishing (Willis & Babcock 2000, Cappo et al. 2004, Watson et al. 2005, Watson et al. 2007). Other research, however, demonstrates that mobile predators can be accurately surveyed using UVC (Friedlander & DeMartini 2002, Castro & Rosa 2005, Robbins et al. 2006). In our research, we investigate differences in how the 2 methods measure abundance and richness of both conspicuous species and mobile predators.

In the detection spectrum, cryptic species are at the opposite end to conspicuous species. These fishes are rarely detected in visual surveys, which are well known to underestimate their density (Willis 2001). In fact, the only way to accurately survey cryptics may be with the application of an ichthyocide (Willis 2001). While the chance of detecting cryptic species is low on a UVC, it is even smaller when using diver-swum underwater video (Watson et al. 2005). When the underwater video is stationary, as in the case of a BRUV, the chance of seeing cryptic species declines further (Stobart et al. 2007). In addition to cryptic colouration and secretive behaviour, cryptic species often have small maximum total lengths (TLmax).

Interestingly, MacNeil et al. (2008b) found TLmax to be negatively associated with detection probability, and proposed that the territorial behaviour of the small-bodied species in their study compensated for their small size, with the result that these small- bodied species were more detectable than larger-bodied transitory species. The effect that territoriality has upon survey methodology has received little attention in the literature. There are good a priori reasons to expect that UVC and BRUV might provide different estimates of the abundance of territorial species. A diver conducting a UVC is likely to pass through many individuals’ territories, while a BRUV unit will usually land in a single individual’s territory or the junction area between a few individuals’ territories. Similar to Willis & Babcock (2000), we frequently observed individuals interacting in an antagonistic manner around the bait. We predicted that the result of these behaviours would limit a territorial species’ ability to respond to the bait, thereby allowing UVC to record higher abundances of these species than BRUV.

- 10 -

Another group of species that might be limited in their response to bait are herbivores. While it is tempting to suppose that BRUV does not survey herbivores at all, this is not the case. Harvey et al. (2007) found that herbivores were attracted to a BRUV more than to an unbaited video unit, with attraction most likely being to the activity around the frame rather than to the bait itself. However, this study did not compare herbivore densities measured using BRUV to those measured using UVC. In our research, we explore differences in abundances measured using UVC and BRUV with the specific prediction that UVC will record higher abundance of herbivores. One of the identified sources of bias using UVC is the difficulty that divers may have in accurately counting individuals (Harvey et al. 2004), which could be especially true if a species forms dense schools. While schooling behaviour increases the chance that a species will be detected (MacNeil et al. 2008b), it is unclear whether UVC can provide an accurate estimate of this group’s abundance. Measures of abundance of these species using BRUV may also be compromised: There is an upper limit to the number of fish that can be viewed in a frame (Willis et al. 2000). Here, we explore differences in these methods’ abilities to record abundance and richness of densely schooling species.

Our methodological comparison utilizes the diverse fish assemblages inhabiting temperate rocky reefs in Southeast Australia. This assemblage has received little scientific investigation and there is a pressing need to establish appropriate survey protocols for (1) assessing the response of this assemblage to marine protected area (MPA) establishment, (2) identifying locations/species of high conservation value, and (3) testing the applicability of the bioregional province framework developed for the management of southern Australian marine waters to reef fishes. We utilize this assemblage to compare performance of these 2 methods in quantifying diversity as well as to explore intrinsic factors that influence a method’s ability to detect a species. No comparison between methods is complete without an assessment of effort. While some research has found BRUV stations to be more efficient in terms of statistical power, personnel hours, and boat resources (Cappo et al. 2003, Watson et al. 2005), each research program will differ in resource availability and application. Here, we compute the amount of time to conduct a single sample of each method in order to compare performance of both methods using a standardized metric.

- 11 -

MATERIALS & METHODS Locations Surveys were conducted on nearshore rocky reefs at 4 locations along the coast of Victoria in Southeast Australia (Fig. 1) between December 2007 and June 2008 (Table 1). Locations were chosen to include MPAs, and to be representative of 3 biogeographic provinces and 1 transition zone (Hough & Mahon 1994). To ensure that surveys occurred in the correct habitat in a patchy matrix of sand and rocky reef, we selected sites based on depth (between 4 and 20 m) and real-time side scan sonar imagery of the benthos. We aimed to conduct 1 UVC for every 3 BRUV units deployed. However, due to a combination of equipment malfunction and poor weather, we were unable to achieve this ratio at all sites. The actual numbers of surveys of each type are provided in Fig. 1.

BRUV To capture underwater video footage, we placed an HC-series Sony Handycam in an Ocean ImagesTM underwater housing that was mounted into a weighted aluminium frame (Cappo 2006). A tube of PVC extended from the frame’s lid, from which we suspended a mesh bait pouch containing 400 g of crushed pilchards, 1.5 m from the lens (Fig. 2). Trial deployments with the bait arm attached beneath the camera were unsuccessful as kelp often occluded the bait bag. The bait pouch rested within ~20 cm of the benthos when full and ~30 cm above when empty. The unit was deployed from a boat onto or immediately next to rocky reef. After deployment, the boat motored away from the area. We deployed 2 BRUV units concurrently at a minimum of 500 m apart to minimize the overlap of bait odour plumes (Willis & Babcock 2000, Harvey et al. 2007, Heagney et al. 2007). Where BRUV units could not be separated spatially, replicates occurring <500 m apart were deployed at least 1.5 h apart.

Preliminary data collected from Barwon Heads suggested that BRUV immersions of 60 min duration were required to record many species in Victorian coastal waters, as 47% of maximum species counts occurred after 30 min, and 24% occurred after 45 min (M. A. Colton unpubl. data). For each species on each tape, we enumerated the maximum number of individuals observed in a single frame (MaxN). MaxN is a conservative measure of relative density that avoids the recounting of individuals that repeatedly visit the bait. A single viewer watched all the tapes at least once on a - 12 -

Figure 2: The BRUV unit. The frame is constructed of aluminium; the bait arm is PVC pipe; the bait pouch is made of gutter mesh; the bridle ropes connect to a surface buoy and allow the unit to be deployed and retrieved remotely; two lead dive weights (1.5 kg each) are attached to each leg; and the video camera is mounted into the underwater housing. computer monitor. Species of questionable identification were flagged and viewed again, and the assistance of taxonomists1 was called upon when necessary. As the viewer’s ability to identify fishes improved over time, the first 12 tapes examined were viewed twice.

UVC To conduct UVCs, an observer and dive buddy on SCUBA descended to the substrate at the site of a previous BRUV deployment (Fig. 1). The time between a BRUV deployment and a subsequent UVC survey varied (Table 1), with a median number of 39 d elapsing between survey types at all locations and a minimum of 1 d at any location. We tested to ensure that differences in the time elapsed between methods did not confound our results by removing from analyses the location with the largest number of days between BRUV and UVC surveys (below). After waiting 2 min for the disturbance of their arrival to dissipate (Samoilys & Carlos 2000), the observer swam slowly in a pre-determined direction, looking ahead as well as under ledges and in the kelp understory, identifying and counting fish in a strip 5 m wide. The dive buddy, present for safety reasons, remained behind the observer. The direction of each census was chosen prior to the dive and was based upon the strength and direction of currents, as well as the ability of the boat to retrieve the divers. The latitude and

1 Dr. Martin F. Gomon, Senior Curator, and Dianne J. Bray, Fish Collections Manager, Museum Victoria, Carlton, Victoria 3053, Australia - 13 -

Table 1: Description of sampling effort by method at each location. Start and end dates and number of days spent sampling are listed for both methods combined. The median number of days that separated the application of the two methods is provided for each location. Distance and depth -1 and are in metres, times are in minutes, and swim speed is in m min . For BRUV, Tsoak is the total time the frame is immersed, and Tdepth is the time the frame spends at depth. All values are mean ± SD. UVC & BRUV UVC BRUV Median # Start End # days b/w

Location Date Date days methods Depth Distance Speed Duration Depth Tsoak Tdepth Apollo 22Jan08 03Mar08 41 34 13.5 ± 4.8 502 ± 300 14 ± 9 34 ± 8 11.2 ± 5.0 59.8 ± 4.0 60.0 ± 2.0 Barwon 19Dec07 17Jun08 181 155 10.6 ± 4.5 261 ± 84 10 ± 2 27 ± 7 10.3 ± 4.7 61.0 ± 2.7 59.2 ± 4.7 Prom 18Mar08 05Jun08 79 70 12.6 ± 4.7 272 ± 70 9 ± 3 30 ± 1 11.7 ± 4.7 61.8 ± 0.4 61.2 ± 0.5 Howe 05Apr08 19Apr08 14 10 14.1 ± 3.9 314 ± 102 9 ± 4 37 ± 7 13.6 ± 4.2 58.3 ± 8.8 56.7 ± 9.0 -

longitude of diver entry and egress points were used to compute the linear transect distance. Though divers made every effort to swim in a straight line using a compass heading, deviations were inevitable, making the linear survey distance an underestimate of the actual area surveyed. However, deviations from the straight line distance will represent only a small proportion of the total survey area because transects were so long (Table 1). The dive time was limited by no-decompression limits in dive tables or by tank duration (Table 1). Swimming speeds were comparable to those in other surveys (see Lincoln Smith 1988 for a review) (Table 1). All observers were highly trained in fish identification, and 80% of UVCs were conducted by the same researcher.

Lincoln Smith (1988) found that differences in observer responses to either time- based or distance based methods can result in biased estimates of some species’ abundances. We chose to conduct UVCs with unregulated distances and times, allowing habitat complexity, the presence of fish, and currents to determine our swimming speed. Like Fowler (1987), we found no relationship between loge- transformed transect length and species richness, diversity, or number of individuals for all locations combined using Pearson product-moment correlations (r = –0.05, r = 0.19, r = –0.06 respectively; all n = 35, p > 0.05) (M. A. Colton unpubl. data). We avoided the use of transect tape in order to more effectively use our bottom time (Patterson et al. 2007), reduce our disturbance (Fowler 1987), and conduct long- distance UVCs without running into problems arising from the deployment of long lengths of tape in considerable surge. Unfortunately, the use of sonar to measure transect lengths (Patterson et al. 2007) was not found effective in the turbid waters which typify our study sites.

Between-methods comparisons using relative abundance We examined differences in species composition using relative abundance, measured as relative density using UVC and relative MaxN using BRUV. For univariate analyses in which the information in a sample was condensed to a single number, e.g. Shannon Diversity, we computed relative abundance as the total number of individuals in a sample divided by the total number of individuals observed in all samples using that method. For multivariate analyses, we computed the relative abundance of a species in a sample by dividing the number of individuals of a species - 15 -

in a sample by the total number of individuals in the sample. Means were computed using all samples only from the location(s) in which a species was observed. Individuals were identified to species wherever possible. There were, however, some differences in the ability of the methods to identify fishes: using BRUV, it was difficult to differentiate between species in the families Diodontidae and Urolophidae, though these could be distinguished using UVC. For comparisons between methods, species observed using UVC from these 2 families were combined into 2 family level groupings.

Diversity and abundance Measures of diversity were compared between the methods using univariate analyses, and mean relative abundances were compared using multivariate analyses. As UVCs were conducted at approximately the same sites in which BRUVs were deployed, we used paired t-tests (SPSS v16.0) to compare measures of diversity, which controlled for differences in habitat that naturally occur in patchy reef environments. Before conducting t-tests, all dependent variables were tested for normality, and a fourth-root transformation was applied to normalize the variances of number of individuals as the data were highly skewed. As mentioned previously, we attempted to always compare 3 BRUV deployments with 1 UVC, though technical difficulties occasionally precluded achievement of this ratio. Therefore, mean values computed from up to 3 BRUV deployments were compared with the value computed from 1 UVC. We examined species richness, family richness, fourth-root transformed number of individuals and diversity as measured by the Shannon Index:

= 푠 ( )ln( ) (1) ′ 퐻 � 푝푖 푝푖 where pi is the proportion of individuals푖=1 in a sample belonging to species i.

Differences in diversity between the methods were also examined using a multivariate approach. We employed the ANOSIM routine in Primer-E (Clarke & Warwick 2001a) to compare between methods including the factor location, using a Bray-Curtis dissimilarity matrix of fourth-root transformed relative abundance. Those species responsible for the observed differences were identified using the SIMPER routine.

- 16 -

As UVCs and BRUVs were conducted within the same season at all locations except Barwon Heads (Table 1), we repeated our univariate and multivariate analyses omitting Barwon Heads, and assessed significance using a Bonferroni-adjusted α- level of 0.03.

Taxonomic distinctness In addition to species richness and diversity, we tested for between-methods differences in taxonomic distinctness. A Linnaean classification system was used including the ranks species, genus, sub-family, family, order, and class. Assignments to species, genus, sub-family and family were based on the Catalog of Fishes (Eschmeyer & Fricke 2009), while the ranks of order and class were obtained from Gomon et al.(2008). Two measures of taxonomic diversity were examined: Δ+, or the average path length between species (Clarke & Warwick 1998); and Λ+, or the variance of the pairwise path lengths (Clarke & Warwick 2001b). We used the TAXDIST routine in PRIMER-E (Clarke & Warwick 2001a) to create funnel plots with 95% confidence intervals for presence–absence data collected using both methods. Equal weights were assigned to all taxonomic ranks; changing the weight for sub-family to half that of the other ranks was found to make no difference to the results.

Intrinsic factors affecting detectability In addition to measures of diversity and richness, we also examined which kinds of species were best recorded by each method. Species were classified into sightability types using an adaptation of the criteria developed by Lincoln Smith (1989). Type I species were classified as site-attached, often small-bodied or cryptic, e.g. Parma microlepis (Pomacentridae). Type II species form dense schools in the water column, e.g. Atypichthys strigatus (Microcanthinae). Type III species are conspicuous, either occurring as individuals or in small groups, e.g. Notolabrus tetricus (Labridae).

In addition to sightability, we tested whether territorial species, herbivores, or mobile predators were better sampled by either method. We hypothesized that herbivorous species would not be attracted to the bait in as high numbers as carnivorous or omnivorous species, and that higher abundances of territorial species would be recorded by UVC as these species could be limited in their ability to respond to bait. - 17 -

Table 2: Results from a SIMPER routine in Primer-E used to identify the species that contribute to differences between methods. Species are listed in order of their contribution. Mean relative abundances, x ± SE, are listed for each method and were computed from only the location(s) in which a species was observed. � Species Family (UVC) (BRUV) % Contribution Cumulative % Chrysophrys auratus Sparidae 0.004 ± 0.002 0.064 ± 0.013 4.5 4.5 퐱� 퐱� Odax cyanomelas Odacidae 0.051 ± 0.008 0.017 ± 0.003 4.0 8.5 Pseudocaranx spp. Carangidae 0.014 ± 0.011 0.180 ± 0.030 3.8 12.3 Pempheris multiradiata Pempheridae 0.052 ± 0.017 0.005 ± 0.004 3.6 15.9 Upeneichthys spp. Mullidae 0.027 ± 0.006 0.032 ± 0.007 3.5 19.4 Notolabrus fucicola Labridae 0.033 ± 0.012 0.023 ± 0.008 3.3 22.8 Acanthaluteres vittiger Monacanthidae 0.029 ± 0.006 0.005 ± 0.002 3.3 26.0 Meuschenia freycineti Monacanthidae 0.008 ± 0.004 0.022 ± 0.004 3.2 29.2 Caesioperca rasor Serranidae 0.084 ± 0.030 0.039 ± 0.011 3.1 32.3 -

18 Pictilabrus laticlavius Labridae 0.015 ± 0.003 0.018 ± 0.004 3.1 35.3

- Meuschenia hippocrepis Monacanthidae 0.077 ± 0.030 0.063 ± 0.009 3.0 38.3

Cheilodactylus nigripes Cheilodactylidae 0.028 ± 0.006 0.021 ± 0.004 3.0 41.3 Dinolestes lewini Dinolestidae 0.054 ± 0.020 0.017 ± 0.006 2.9 44.3 Enoplosus armatus Enoplosidae 0.017 ± 0.006 0.011 ± 0.005 2.9 47.2 Scorpis aequipinnis Kyphosidae 0.045 ± 0.013 0.013 ± 0.005 2.8 49.9 Girella zebra Kyphosidae 0.028 ± 0.017 0.009 ± 0.004 2.7 52.6

- 18 -

We also predicted that large, mobile predators, such as many elasmobranchs, would be better sampled by BRUV than UVC, as has been shown by some other studies (Willis & Babcock 2000, Cappo et al. 2004, Watson et al. 2005, Watson & Harvey 2007). We identified species belonging to the 3 groups using data from guidebooks (Kuiter 2000, Edgar 2005, Gomon et al. 2008). We used independent samples t-tests, with degrees of freedom computed using the Welch-Satterthwaite formula in SPSS (v 16.0) to correct for unequal sample sizes and heteroscedasticity, to examine whether the methods differed in the mean number of species ([ln (x + 1)]-transformed) of each sightability type. We also tested whether the relative abundance (fourth-root transformed) of species of the different sightability types and groups (herbivores, territorial species, and mobile predators) differed between methods. We used a Bonferroni-adjusted α-level of 0.03 to assess significance.

Between methods comparison of effort We compared effort between methods firstly by using species accumulation curves to estimate the maximum species richness for each method, and secondly by computing the number of samples required by each method to observe a proportion of the maximum species richness. Finally, we computed a metric that standardized these samples, allowing us to compute the amount of effort, measured in time, required by each method to achieve a proportion of maximum species richness.

Species accumulation curves were constructed using EstimateS (v. 8.00, Colwell 2006) for UVC and BRUV separately at each location. Species accumulation curves assume that replicate samples are equal. Although oceanographic conditions, bait dispersal and therefore the area surveyed by a BRUV unit will not always be the same among deployments, and individual UVC transects also sampled different areas (Table 1), we make the simplifying assumption that BRUV deployments are equal and that one UVC sample is equivalent to the mean distance covered in all transects (mean ± SD = 360 ± 212 m). However, as EstimateS randomises the ordering of samples during the simulations, such potential variation among samples will simply widen the confidence intervals of the estimate.

To estimate the maximum species richness observable for each method, we used NLREG (Sherrod 2008) to fit the Michaelis-Menten function (Raaijmakers 1987) - 19 -

using Sobs (Mao τ), the observed species richness calculated by EstimateS, as a measure of S(x), which is the number of species, S, observed at a given level of sampling effort, x: ( ) ( ) = (2) + 푥 푆푚푎푥 where b and S are constants,푆 푥and S is also the predicted asymptotic species max max푏 푥 richness. We assessed goodness of fit using r2 values calculated by NLREG, all of which were >0.95.

To determine the number of samples required to obtain a certain percentage of

predicted species richness, we solved Eq. (2) for x given y = Z(Smax), where Z is a

proportion of the predicted species richness (Smax):

= (3) 1 −푍푏 We then computed x given Z = 60%,푥 75%, and 95% for each method and location. 푍 −

Finally, we determined how much time it took to complete a single sample using each method, and utilized these values to compute a standardization metric. To complete a sample, each method requires time in the field and in the lab, and, for our research program, the methods required the same number of personnel in both settings. As the parameters required to convert time to cost will be specific to a research program, here we only report the time taken in our surveys to deploy a single BRUV unit and conduct 1 UVC transect. Our research used 2 BRUV units, which we deployed from a small vessel (6 m length).

We computed the mean field-time to deploy a single BRUV unit from 14 sampling days in which we only conducted BRUV surveys. To each day of sampling, we added 20 min per camera to account for the time to set up and break down the equipment. The mean time to launch and retrieve a single BRUV unit, including time spent recording, was 54 field-min. There is a well known bottleneck (Cappo et al. 2003) in the analysis of BRUV tapes, with the time to analyse a tape dependent upon the number of individuals on the tape and the ease by which species can be identified. Estimates of time spent in the lab on BRUV analysis range from 2× the length of the tape (Willis et al. 2000) to 24× the time spent on a UVC (Stobart et al. 2007). In our

- 20 -

25 a) BRUV: above 50th percentile ) 1 – - 20

5 Mean MaxN (deployment

0 § § § § § § § § § § § § -

21 Atherinidae Atherinidae

(M) (M) Arripis spp. (H) Girella zebra zebra Girella † (H) Meuschenia galii † Meuschenia Scorpis lineolata Scorpis Upeneichthys spp. Upeneichthys † Caesioperca rasor rasor Caesioperca Pseudocaranx spp. Pseudocaranx Notolabrus fucicola Notolabrus † (M) Pagrus † (M) auratus Enoplosus armatus † Chromis hypsilepis hypsilepis Chromis Meuschenia scaber † Meuschenia Scorpis Scorpis aequipinnis Pictilabrus laticlavius † laticlavius Pictilabrus (T) (T) Parma microlepis ‡ Atypichthys strigatus strigatus Atypichthys (M) Dinolestes lewini lewini Dinolestes (M) Acanthistius ocellatus Acanthistius † Meuschenia freycineti † freycineti Meuschenia (H, T) (H, T) Parma victoriae‡ (T) (T) tetricus Notolabrus † (M) Trygonorrhina spp.† Trygonorrhina (M) Cheilodactylus nigripes Cheilodactylus † Notolabrus gymnogenis gymnogenis Notolabrus † Nemadactylus douglasii Nemadactylus † Meuschenia hippocrepis † Meuschenia (H, (H, T) Odax † cyanomelas Ophthalmolepis lineolata † Ophthalmolepis Caesioperca lepidoptera lepidoptera Caesioperca Eubalichthys bucephalus Eubalichthys † Parequula † Parequula melbournensis (M) (M) barbata Pseudophycis † (M) (M) Sillaginodes punctatus † (M) Gymnothorax prasinus ‡ prasinus Gymnothorax (M) Trachurus novaezelandiae novaezelandiae Trachurus Trachinops Trachinops caudimaculatus Hypoplectrodes maccullochi ‡ Hypoplectrodes (M) (M) laticepsCephaloscyllium †

(M) (M) Heterodontus portusjacksoni † Figure 3 (This and following pages): Mean abundance measured by (a) & (b) mean MaxN for BRUV samples, and (c) & (d) mean density (m–2) for UVC samples. Species shown in (a) & (c) are those above the 50th percentile of abundance, and (b) & (d) are those below the 50th percentile of abundance. Means were computed only from location(s) in which species occurred. Sightability types (see ‘Materials and Methods’) are shown after species’ names, with: ‡ = Type I, § = Type II, † = Type III. Letters in parentheses at the start of a species’ name indicate group, with: H = herbivores, T = territorial species, M = mobile predators. Black bars = species unique to a single method; white bars = species from families that could be identified to species level using UVC but not BRUV; grey bars = species that could be identified to species level and that were observed by both methods. Error bars are + SE.

- 21 -

0.4 b) BRUV: below 50th percentile ) 1 – - 0.3

0.2

0.1 Mean MaxN (Deployment

0.0 -

Diodontidae † Diodontidae -

Aracana aurita aurita Aracana † Cyttus australis † Cyttus Odax Odax acroptilus † (M) (M) † Urolophidae (M) † lineata Latris (M) (M) lalandi † Seriola (M) Eeyorius Eeyorius hutchinsi † Achoerodus viridis viridis Achoerodus † Tetractenos glaber glaber † Tetractenos (M) spp.† Asymbolis (M) Haletta Haletta semifasciata † (M) Dipturus whitleyi † whitleyi Dipturus (M) (M) (M) spp. Orectolobus † Tilodon sexfasciatum † Tilodon sexfasciatum Meuschenia australis † Meuschenia (M) Conger verreauxi ‡ verreauxi Conger (M) Cheilodactylus fuscus Cheilodactylus † Aulopus purpurissatus † (M) (M) Latridopsis forsteri † (H) (H) Crinodus lophodon † Eubalichthys mosaicusEubalichthys † nigricans † Dactylophora (M) Myliobatis australis† Myliobatis (M) Pempheris multiradiata multiradiata Pempheris ‡ Paraplesiops meleagrisParaplesiops † Meuschenia flavolineata † flavolineata Meuschenia (M) (M) Mustelus antarcticus † Chironemus marmoratus marmoratus Chironemus † Pseudolabrus psittaculus Pseudolabrus † Eupetrichthys angustipes Eupetrichthys ‡ Cheilodactylus spectabilis Cheilodactylus † (H) Acanthaluteres vittiger † vittiger Acanthaluteres (H) (M) (M) Dasyatis † brevicaudata (H) (H) arctidens † Aplodactylus Nemadactylus macropterus macropterus Nemadactylus † Pentaceropsis recurvirostris † recurvirostris Pentaceropsis Neosebastes scorpaenoides † Neosebastes scorpaenoides (M) (M) Platycephalus bassensis † (M) (M) Notorynchus cepedianus †

(M) (M) Sphyraena † novaehollandiae Figure 3 Continued

- 22 -

0.12 c) UVC: above 50th percentile

0.1 ) 2 – 0.08

0.06

0.04 Mean density Mean density (m 0.02

0 § § § § § § § § § § -

(H) Girella zebra zebra Girella † (H) Scorpis lineolata Scorpis Achoerodus viridis viridis Achoerodus † Upeneichthys spp. Upeneichthys † Caesioperca rasor rasor Caesioperca Notolabrus fucicola Notolabrus † Enoplosus armatus † Chromis hypsilepis hypsilepis Chromis Meuschenia scaber † Meuschenia Pseudocaranx spp. Pseudocaranx Scorpis Scorpis aequipinnis (M) Dinolestes lewini † lewini Dinolestes (M) Pictilabrus laticlavius † laticlavius Pictilabrus (T) (T) Parma microlepis ‡ Atypichthys strigatus strigatus Atypichthys Trachinops taeniatus Trachinops Cheilodactylus fuscus Cheilodactylus † (H, T) (H, T) Parma victoriae‡ (T) (T) tetricus Notolabrus † Pempheris compressaPempheris ‡ (H) (H) Crinodus lophodon † Pempheris multiradiata multiradiata Pempheris ‡ Cheilodactylus nigripes Cheilodactylus † Notolabrus gymnogenis gymnogenis Notolabrus † Nemadactylus douglasii Nemadactylus † Meuschenia flavolineata † flavolineata Meuschenia Meuschenia hippocrepis † Meuschenia (H, (H, T) Odax † cyanomelas Ophthalmolepis lineolata † Ophthalmolepis (M) Chrysophrys † auratus Chrysophrys (M) Chironemus marmoratus marmoratus Chironemus † Caesioperca lepidoptera lepidoptera Caesioperca Eupetrichthys angustipes Eupetrichthys ‡ Parequula † Parequula melbournensis Cheilodactylus spectabilis Cheilodactylus † (H) Acanthaluteres vittiger † vittiger Acanthaluteres (H) (M) (M) Sillaginodes punctatus † Trachurus novaezelandiae novaezelandiae Trachurus Siphonognathus beddomei Siphonognathus ‡ Trachinops Trachinops caudimaculatus Hypoplectrodes maccullochi ‡ Hypoplectrodes

(M) (M) Heterodontus portusjacksoni † Figure 3 Continued

- 23 -

d) UVC: below 50th percentile 0.0012

0.001 ) 2 – 0.0008

0.0006

0.0004 Mean Density (m 0.0002

0 … § § § § -

- Atherinidae Atherinidae Apogonidae Apogonidae (M) Sphyraena Sphyraena (M)

Girella elevata † elevata Girella Aracana aurita aurita Aracana † Lotella rhacina † Aracana ornata ornata Aracana † Odax Odax acroptilus † Coris Coris sandageri † Optivus agastos ‡ Optivus Meuschenia galii † Meuschenia Pempheris affinis Pempheris ‡ Trygonoptera spp.† Trygonoptera Tetractenos glaber glaber † Tetractenos Neoodax balteatus † Neoodax Myliobatis australis † Myliobatis Suezichthys aylingi aylingi † Suezichthys Urolophus cruciatus † Urolophus Diodon nicthemerus † Diodon Haletta Haletta semifasciata † Meuschenia venusta venusta Meuschenia † vanessa Lepidotrigla † Callanthias australis Callanthias (T) (T) Parma microlepis ‡ Tilodon sexfasciatum † Tilodon sexfasciatum Meuschenia australis † Meuschenia Acanthistius ocellatus Acanthistius † Thamnoconus degeni Thamnoconus † Meuschenia freycineti † freycineti Meuschenia (M) (M) Latridopsis forsteri † Trachichthys australis Trachichthys (M) Trygonorrhina spp.† Trygonorrhina (M) Dactylophora nigricans † Dactylophora Dotolabrus aurantiacus Dotolabrus ‡ Phyllopteryx taeniolatus ‡ Phyllopteryx (M) (M) ornatus Orectolobus † Eubalichthys bucephalus Eubalichthys † Pseudolabrus psittaculus Pseudolabrus † Scobinichthys granulatus granulatus Scobinichthys † Dicotylichthys punctulatus † Dicotylichthys (H) (H) arctidens † Aplodactylus (M) Gymnothorax prasinus ‡ prasinus Gymnothorax (M) Siphonognathus attenuatus attenuatus ‡ Siphonognathus Pentaceropsis recurvirostris † recurvirostris Pentaceropsis (M) (M) laticepsCephaloscyllium †

Figure 3 Continued

- 24 -

computations, we use an estimate of 2× BRUV tape length as a measure of minimum analysis time. In addition, we speculate that it takes ~20 min to enter the data collected from the tape into an appropriate computer program. Therefore, the total mean time to deploy, retrieve and analyse a single BRUV tape is ~190 min.

To compute the time spent conducting a single UVC, we took the mean survey time plus 15 min spent gearing up at the start of a dive, 8 additional min in the water (1 min to reach the benthos, a 2 min wait period at the start of the dive, 2 min to the surface, and 3 min on a safety stop), and 5 min getting onto the boat at the end of a dive. Lab time for UVC is minimal: we speculated that it takes ~20 min to enter the data into an appropriate computer program. This resulted in an average of ~80 min for 1 UVC. Using these times, we computed that we could complete about 2 UVC transects for every 1 BRUV unit deployed. To determine the amount of time required to obtain the same percentages of total species richness for each method, we multiplied the number of BRUV samples by 2 to standardize for the between-method differences in the time taken to obtain a sample.

RESULTS We observed a total of 78 species belonging to 44 families using BRUV, and 85 species belonging to 42 families using UVC (Fig. 3). Seventeen species were only observed by BRUV, and 20 species were only observed by UVC. Of those species that were frequently observed, i.e. within the upper 50th percentile, only 3 were exclusively observed using UVC and only 2 were exclusively observed using BRUV (Fig. 3). The 3 families occurring most frequently in BRUV samples were Labridae (observed in 89% of the samples), Monacanthidae (66%), and Carangidae (58%). The 3 families that occurred most frequently in UVC samples were Labridae (100% of samples), Monacanthidae (100%), and Cheilodactylidae (89%).

Diversity and abundance Using paired t-tests, we found that UVC recorded significantly higher species diversity (t(34) = 3.66, p = 0.001), species richness (t(34) = 7.19, p < 0.0005), family richness (t(34) = 6.30, p < 0.0005), and number of individuals (t(34) = 10.94, p < 0.0005) than BRUV when all locations were examined (Fig. 4) and when the Barwon

Heads site was excluded from the analysis (Shannon Index diversity t(27) = 3.11, p = - 25 -

25 5 BRUV UVC ** 20 ** 4 Mean # individuals # Mean

15 ** 3

10 2 Mean diversity and richness Meandiversity 5 1 *

0 0 Shannon Diversity # Species # Families #Individuals

Figure 4: Paired t-tests comparing diversity and abundance between BRUV and UVC for all locations. Asterisks indicate significance: *p = 0.001; **p < 0.0005. Error bars are ± SE.

0.004; species richness t(27) = 4.33, p < 0.0005; family richness t(27) = 3.8, p = 0.001; and number of individuals t(27) = 11.551, p < 0.0005).

Using the ANOSIM routine in Primer-E (Clarke & Warwick 2001a), we found a significant difference between methods in the relative abundance of species for all locations (R = 0.225, p < 0.001) (Fig. 5), and when Barwon Heads was excluded (R = 0.218, p < 0.004). Using the SIMPER routine, we found that 16 species were required to explain >50% of the difference between methods, and that no single species contributed >4.5% (Table 2). Notably, the average abundances of the top 4 species differed between BRUV and UVC, with half observed in higher numbers by BRUV, and half by UVC.

Taxonomic distinctness We examined how the methods compared in 2 measures of taxonomic distinctness, Δ+ and Λ+, using the TAXDIST routine in Primer-E (Clarke & Warwick 2001a). UVC was found to survey a less taxonomically distinct population, i.e. to have lower Δ+, than BRUV (Fig. 6a). However, when the evenness of the taxonomic tree, i.e. Λ+, was taken into account, there was little difference between methods (Fig. 6b).

- 26 -

Table 4: Species accumulation curves were used to determine the predicted maximum species richness, Smax, for each method at each location by fitting a Michaelis-Menten function (Eqn. 2). The number of samples required by each method to achieve a proportion, Z, of Smax (Eqn. 3) is listed for each method at each location. Note that these numbers have not been standardized by the relative amount of time required to complete a sample of each method.

Number of samples

Smax ± SE Z = 60% Z = 75% Z = 95% Location BRUV UVC BRUV UVC BRUV UVC BRUV UVC Apollo 46 ± 0.6 53 ±0.9 7 4 14 8 90 53 Barwon 44 ± 0.8 43 ±1.2 6 4 13 9 82 56 Prom 62 ± 0.8 56 ± 1.8 11 3 21 6 135 39 Howe 58 ± 0.2 71 ± 0.8 6 2 13 5 82 31

- 27 -

Table 3: Results from independent samples t-tests, with degrees of freedom computed using the Satterthwatite-Welch formula, comparing between methods the 4th-root transformed mean relative abundance of mobile predators. Significance was assessed at α-level = 0.03; significant p-values and species are in bold.

Family Species (BRUV) ± SE (UVC) ± SE t (df) p Arripidae Arripis spp. 0.02 ± 0.02 0 1.0 (21) 0.329 Carangidae Seriola lalandi 0.03퐱� ± 0.02 퐱� 0 1.4 (23) 0.162 Congridae Conger verreauxi 0.02 ± 0.02 0 1.0 (21) 0.329 Dasyatidae Dasyatis brevicaudata 0.06 ± 0.02 0 3.6 (84) <0.0005 Dinolestidae Dinolestes lewini 0.15 ± 0.03 0.19 ± 0.05 -0.7 (61) 0.467 Heterodontidae Heterodontus portusjacksoni 0.12 ± 0.02 0.03 ± 0.01 3.6 (118) 0.001 Hexanchidae Notorynchus cepedianus 0.02 ± 0.02 0 1.0 (21) 0.329 Latridae Latridopsis forsteri 0.02 ± 0.01 0.09 ± 0.04 -1.7 (32) 0.092 Latridae Latris lineata 0.01 ± 0.01 0 1.0 (21) 0.329 Moridae Pseudophycis barbata 0.07 ± 0.02 0 3.0 (44) 0.004 Muraenidae Gymnothorax prasinus 0.20 ± 0.04 0.05 ± 0.03 2.9 (24) 0.008 -

Myliobatidae Myliobatis australis 0.10 ± 0.02 0.02 ± 0.01 3.6 (114) <0.0005 28 Orectolobidae Orectolobus spp. 0.01 ± 0.01 0.02 ± 0.02 -0.4 (11) 0.705

- Platycephalidae Platycephalus bassensis 0.05 ± 0.02 0 2.0 (44) 0.050 Rajidae Dipturus whitleyi 0.02 ± 0.01 0 1.7 (67) 0.103 Rhinobatidae Trygonorhina spp. 0.10 ± 0.02 0.05 ± 0.02 1.6 (77) 0.120 Scyliorhinidae Asymbolis spp. 0.01 ± 0.01 0 1.0 (23) 0.328 Scyliorhinidae Cephaloscyllium laticeps 0.14 ±0.02 0.07 ± 0.02 2.0 (103) 0.047 Sillaginidae Sillaginodes punctatus 0.15 ± 0.03 0.02 ± 0.02 3.3 (84) 0.002 Sparidae Chrysophrys auratus 0.28 ± 0.03 0.05 ± 0.02 6.1 (117) <0.0005 Sphyraenidae Sphyraena novaehollandiae 0.04 ± 0.01 0.01 ± 0.01 2.2 (90) 0.031 Triakidae Mustelus antarcticus 0.03 ± 0.02 0 1.4 (23) 0.170 Urolophidae Various spp. 0.02 ± 0.01 0.05 ± 0.03 -1.0 (38) 0.303

- 28 -

Intrinsic factors affecting detectability Both methods were good at detecting conspicuous (Type III) species and poor at detecting cryptic (Type I) species. Of all the species observed, 73% were classified as Type III, 15% as Type I and 12% as Type II (Fig. 3). Of those species observed only using BRUV, 6% were Type I, 6% were Type II, and 88% were Type III. Of the species observed only using UVC, 41% were Type I, 9% were Type II, and 50% were Type III.

Using UVC we recorded significantly more Type I and Type III species (t(62) = 7.602 and t(111) = 6.065, respectively, both p < 0.0005) and higher relative abundance of

Type I species (t(70) = 7.888, p < 0.0005) than using BRUV (Fig. 7). There was no difference between the methods in terms of relative abundance of Type II and Type

III species (t(73) = –0.325 and t(95) = 1.211, respectively) or in the number of Type II species (t(57) = 1.747) (Fig. 8).

Using independent samples t-tests with the degrees of freedom computed using the Welch-Satterthwaite formula, we found that UVC recorded significantly higher relative abundance of all herbivorous species (Fig. 8a), and 3 out of 4 territorial species (Fig. 8b). In comparison, BRUV observed significantly higher abundance of 7 of the 23 species of mobile predators (Fig. 8c, Table 3).

Comparison of effort We constructed species accumulation curves using EstimateS (Colwell 2006) and then used Eq. (2) to compute Smax for each method at each location (Table 4). At all locations but Wilsons Promontory, Smax was higher using UVC than using BRUV.

However, the differences in Smax were negligible at all locations except Cape Howe where predicted richness using UVC was 13 species higher than that predicted using BRUV. Using Eq. (3), we found that more UVC than BRUV samples were required to observe the same proportion of maximum species richness (Table 4), even without standardizing by the amount of time required to complete a sample of each method. When we standardized using the ratio of 2 UVCs completed in the time to deploy 1 BRUV, the difference in the number of samples required to observe a proportion of species richness grew. For example, at Barwon Heads, it would take 14 BRUV and 8 UVC samples to reach 60% of maximum species richness, and 180 BRUV and 106 - 29 -

Figure 5: MDS plot of ANOSIM testing differences between methods including factor location. Numbers represent locations: 1 = Apollo Bay, 2 = Barwon Heads, 3 = Wilsons Promontory and 4 = Cape Howe.

UVC samples to observe 95% of predicted species richness.

DISCUSSION This study represents one of the first assessments of how 2 methods, UVC and BRUV, compare in surveying subtidal fish assemblages. Unlike other comparisons between these methods (Willis & Babcock 2000, Willis et al. 2000, Watson et al. 2005), we found that UVC recorded more individuals and higher species diversity as measured by species richness, family richness and the Shannon Index (Fig. 4). At each location, we conducted up to 3 times more BRUV deployments than UVCs (Table 1). Based on these numbers alone, we would expect to record higher species diversity using BRUV than UVC. The fact that we found the opposite to be true suggests that these results may be conservative. Though we found UVC to record higher species richness, family richness, Shannon Index diversity and number of individuals, we determined that BRUV was the better method for observing mobile predators and species targeted by fishing. In addition, BRUV recorded a higher taxonomic diversity than UVC (Fig. 6a), though this was not evident when the evenness of the taxonomic tree was taken into account (Fig. 6b). The higher - 30 -

Figure 6: Differences between methods in 2 measures of taxonomic diversity: (a) taxonomic distinctness Δ+, and (b) variation in taxonomic distinctness Λ+. Solid lines = 95% confidence intervals; dashed lines = mean.

- 31 -

taxonomic distinctness recorded using BRUV is most likely attributable to the higher species richness of elasmobranchs observed using this method. The difference between Δ+ and Λ+ can be explained by the fact that the species that contributed to Δ+, e.g. Notorynchus cepedianus, were observed relatively infrequently using BRUV.

The differences between our results and those reported by other studies (Willis & Babcock 2000, Willis et al. 2000, Watson et al. 2005) are likely to be, in part, a consequence of different approaches. Firstly, most of the previous studies examined the relative effectiveness of the methods for surveying only a few select species (Willis & Babcock 2000, Willis et al. 2000). Only one study has compared the relative merits of BRUV and UVC as tools for surveying a fish community (Stobart et al. 2007). Using qualitative methods, Stobart et al. (2007) found that UVC recorded higher diversity and abundance of many species than BRUV, though these results were not statistically significant. The authors suggested that UVC was the better method especially considering its capacity to provide measures of absolute density.

The second and perhaps more important issue complicating comparisons is that of sampling area. For example, previous comparisons between UVC and BRUV have contrasted 30 min soak times with UVC transects of 125 m2 (Willis & Babcock 2000, Willis et al. 2000). A BRUV unit would sample 314 m2 if the odour only dispersed 10 m from the bait in 30 min (Willis et al. 2000), which is most likely a significant underestimate (Willis et al. 2000, Harvey et al. 2007, Heagney et al. 2007). While models for estimating bait odour plume dispersal exist in deep waters (Sainte-Marie & Hargrave 1987, Heagney et al. 2007), they have yet to be developed for turbulent environments leaving estimates of odour dispersal to speculation. However, like other authors (Willis et al. 2000), we feel it is safe to assume that a comparison between a 125 m2 UVC and a 30 min BRUV deployment is a comparison between 2 different sampling areas; i.e. the BRUV is sampling a larger area than the UVC to which it is compared. The mean area surveyed in this study using UVC was 1790 m2, suggesting that our comparison may not be as compromised by differences in sample area as previous studies. However, as the area sampled by BRUV remains unknown, it is possible that our UVC surveys sampled a larger area than that sampled by the BRUV units, leading to the observed differences between methods.

- 32 -

Between-methods comparisons are inherently confounded by differences in the data collected. To standardize our measures of abundance, we conducted comparisons between relative abundance, i.e. proportions. Proportions are dependent upon the total number of individuals observed: The method that records more total individuals will record a lower proportion of individuals of a given species even if the same number of individuals is observed using both methods. If UVC had recorded fewer individuals in total, we would expect the proportion of a given species to be higher than that observed using BRUV. However, as this was not the case (Fig. 4), we are confident that those comparisons for which UVC observed higher relative abundance are valid. While the use of relative abundance will solve some of the complications arising from between-methods comparisons, issues remain. In particular, the use of MaxN, which is a conservative measure, will by definition underestimate abundance. The use of MaxN could therefore contribute to the lower abundance measured using BRUV.

It could be argued that we surveyed more habitat types using UVC than BRUV, resulting in higher species diversity for the method that encountered higher habitat diversity. However, the fact that we completed up to 3 times more BRUVs than UVC transects may serve to counter some of these concerns. Both types of surveys were conducted on rocky reefs, a patchy environment consisting of rocky outcrops separated by sand gullies, and both methods sampled similar habitats including the sand-reef interface and the reef itself. For differences in the number of habitats surveyed to fully explain our results requires that individuals be unable or unwilling to move between habitats in response to bait. While this may be true for some species, we did observe reef-associated species over sand on some videos, suggesting that BRUV may survey more than just the habitat into which it is deployed. However, we also found that UVC recorded higher relative abundance of 3 of 4 territorial species (Fig. 9b), suggesting that site fidelity may account for some of the differences between methods.

Several studies have found that divers are better than cameras at observing cryptic (Type I) species because divers are able to search complex habitats in ways that cameras cannot (Watson et al. 2005, Stobart et al. 2007). Using UVC, we observed more Type I species and more individuals of Type I species than using BRUV (Fig. 7a,b). Similarly, many of the species that were observed only using UVC were Type I, - 33 -

BRUV 16 a) * UVC 14

6 (number of species + 1)

e 4 *

Log 2

0 0.7 b)

0.6

0.5

0.4

0.3 abundance 0.2 * root transformed relative relative transformed root th

4 0.1

0.0 I II III Species Type

Figure 7: Comparison between UVC and BRUV of sightability types (see ‘Materials and Methods’) for (a) numbers of species, and (b) relative numbers of individuals. Asterisks indicate significant differences at p < 0.0005. Error bars are ± SE. suggesting that at least some of the differences between methods can be explained by the better ability of divers to record Type I species. Overall, however, relatively few Type I species were recorded by either method, indicating that neither UVC nor BRUV is particularly effective at observing cryptic species. We estimate that only 38% of the species that could occur in the region were observed in our sampling (M. A. Colton unpubl. data). Like other studies (e.g. Willis 2001, Watson et al. 2005, Stobart et al. 2007), this underscores the need for additional types of survey methods, e.g. ichthyocides, if an adequate sampling of Type I species is desired.

- 34 -

The presence of Type II species could theoretically lead to the type of between- methods differences we observed. BRUV can underestimate density of these species because there is an upper limit to the number of fish that can be viewed in a frame (Willis et al. 2000). However, for only a few BRUV tapes at one location, Cape Howe, was the field of view completely filled with individuals of schooling species (e.g. Atypichthys strigatus, Caesioperca spp.). In contrast, UVC could overestimate density of these species through the recounting of individuals. While we made every effort to avoid this, some individuals of schooling species that appeared to be attracted to divers (e.g. Scorpis spp. and Caesioperca spp.) were probably counted multiple times. However, only 2 Type II schooling species were identified by the SIMPER routine as contributing to the differences between methods: Caesioperca rasor and Scorpis aequipinnis (Table 2). As expected, both of these species were observed in higher numbers by UVC than BRUV. However, they contributed relatively little to the differences between methods: 3.1% and 2.8%, respectively. Indeed, only 12% of all species observed were Type II species. In addition, an independent samples t-test found no difference between methods in the number or abundance of Type II species (Fig. 7), suggesting that densely schooling species cannot wholly account for differences between methods.

We found no significant difference between the methods in the relative abundance of Type II or III individuals (Fig. 7a). However, we found that UVC recorded significantly more Type III species than BRUV (Fig. 7a). As the majority of species observed were Type III (Fig. 3), and UVC observed more species than BRUV, it seems that the majority of differences between the 2 methods can be explained by differences in their ability to record Type III species. Indeed, of the species identified by SIMPER to explain differences between methods, the 2 contributing the most were both Type III, as were 75% of the species contributing to 50% of the between-method variation (Table 2).

One kind of Type III species for which there was a significant difference between methods was mobile predators. BRUV recorded higher mean relative abundance for 7 species of mobile predators, and UVC for none (Table 3). Some of the species better recorded by BRUV are important components of commercial fisheries, e.g.

- 35 -

0.45 a) Herbivores BRUV 0.4 *** UVC 0.35 0.3 ** ** 0.25 ** * 0.2 * 0.15 root relative abundance relative root - 0.1

4th 0.05 0 Acanthaluteres Aplodactylus Crinodus Girella Odax Parma vittiger arctidens lophodon zebra cyanomelas victoriae t(44) = -4.9 t(51) = -2.5 t(7) = -4.0 t(49) = -2.7 t(67) = -5.2 t(51) = -3.6

0.7 b) Territorial species 0.6

0.5 *** *** 0.4

0.3 **

0.2 root mean relative abundance relative mean root

- 0.1 th 4 0 Notolabrus Odax Parma Parma tetricus cyanomelas microlepis victoriae t = -1.7 t = -5.2 t = -6.4 t = -3.6 (117) (68) (29) (51)

0.35 c) Mobile predators 0.3 ***

0.25 * 0.2 * 0.15 ** *** 0.1 *

root relative abundance relative root ** - th 4 0.05

0 Chrysophrys Dasyatis GymnothoraxHeterodontus Myliobatis Pseudophycis Sillaginodes auratus brevicaudata prasinus portusjacksoni australis barbata punctatus

Figure 8: Independent samples t-test results comparing mean relative abundance between methods for (a) herbivores, (b) territorial species, and (c) only those mobile predators for which a significant difference was found. (a) & (b) results from t-tests given as t(df) below species’ names; results for (c) are listed in Table 3. Asterisks indicate significance: *p < 0.03; **p < 0.001; ***p < 0.0005. Error bars are ± SE. - 36 -

Figure 9: Species accumulation curves for each method at each location. One BRUV sample = one BRUV deployment, and on average one UVC sample ≈ 360 × 5 m strip transect. Dashed lines = 95% confidence intervals. Solid lines = Michaelis-Menten functions described as (a) y(BRUV) = (46.2x) (x + 4.8)–1 and y(UVC) = (53.1x) (x + 2.8)–1; (b) y(BRUV) = (43.9x) (x + 4.3)–1 and y(UVC) = (43.5x) (x + 3.0)–1; (c) y(BRUV) = (62.5x) (x + 7.1)–1 and y(UVC) = (56.2x) (x + 2.1)–1; (d) y(BRUV) = (58.1x) (x + 4.3)–1 and y(UVC) = (71.3x) (x + 1.6)–1. Predicted species richness (Smax) is listed for each method at each location.

- 37 -

Chrysophrys auratus and Sillaginodes punctatus. At least 2 other studies (Willis & Babcock 2000, Watson et al. 2005) also found that BRUV was better at observing targeted species. In addition, 5 species of elasmobranch were only observed using BRUV (Fig. 3), which is not surprising as diver-avoidance behaviour by these species has been documented elsewhere (e.g. Watson & Harvey 2007). However, some researchers have found UVC to be an effective method with which to survey mobile predators (e.g. Friedlander & DeMartini 2002, Castro & Rosa 2005, Robbins et al. 2006). In using UVC to survey mobile predators, it may be important to sample larger width and length transects than were used in this research (e.g. Castro & Rosa 2005, Robbins et al. 2006). However, Friedlander & DeMartini (2002) surveyed smaller belt transects, both in terms of length and width, suggesting that the differences we observed between methods may not be wholly attributable to our UVC survey methodology. Indeed, in none of these studies were the densities measured using UVC compared to densities measured using BRUV. It is possible, as our results seem to suggest, that these studies underestimate the densities of elasmobranchs and that another method, such as BRUV, could have recorded higher densities.

Though Harvey et al. (2007) found that BRUV recorded herbivorous as well as carnivorous species, differences in species’ attraction to bait could explain the differences between the methods. There are few data available about dietary preferences for the majority of species observed in this study; we were only able to identify 6 herbivores that were observed more than once during our surveys (Fig. 3). All 6 herbivores were better observed by UVC than BRUV (Fig. 8a), suggesting that dietary preference may explain at least part of the difference between these methods.

We also investigated whether territoriality could explain the differences between UVC and BRUV. For 3 of the 4 species identified as territorial, UVC observed more individuals than did BRUV (Fig. 8b). Two of these species are also herbivorous: the odacid Odax cyanomelas and the pomacentrid Parma victoriae. Their dietary ambivalence towards the bait combined with their territoriality may explain why more individuals of these species were observed by UVC than by BRUV. The 2 pomacentrids, P. microlepis and P. victoriae, which were better observed by UVC, are Type I species which were most often observed under ledges or at the entrance to caves. The inability of stationary cameras to observe these species in complex habitats - 38 -

offers an alternate explanation to between-methods differences. The single territorial species for which there was no significant difference in relative abundance was the labrid Notolabrus tetricus. This species is protogynous and haremic, with only the terminal phase displaying intra-specific antagonistic behaviour around the bait (M. A. Colton pers. obs.). Our inability to separate between phases in our data may mask a true between-methods difference in the relative abundance of this species. To definitively understand whether territoriality influences estimates of abundance recorded by these methods it will be necessary to repeat these comparisons in areas where more known territorial species occur.

In addition to investigating species-specific traits that cause differences between methods, we examined effort. At 2 locations, Wilsons Promontory and Cape Howe, the 95% confidence intervals around species accumulation curves showed no overlap between BRUV and UVC, indicating that for a given number of non-standardized samples, UVC records more species than BRUV (Fig. 9). At the other 2 locations, Apollo Bay and Barwon Heads, the trend is the same. Based on the species

accumulation curves, we computed Smax and found that at 3 locations Smax was higher for UVC than BRUV (Table 4). However, the species accumulation curves, and

therefore Smax values, underestimate differences between methods by treating the samples of each method as equal. In our research, we computed that we could conduct twice as many UVCs as BRUVs in a given unit of time. When we standardized by

time the number of samples required to achieve a proportion of Smax, we found that many more BRUV than UVC samples were required. Therefore, in our research UVC more efficiently sampled species richness.

One of the biggest advantages in the use of BRUV is that units can be deployed in depths and locations that are inaccessible to SCUBA divers and at all times of day and night. Use of BRUV avoids the health and safety concerns associated with SCUBA and may allow more samples to be taken in a given unit of field time. Some research programs report that fewer personnel are required to launch and retrieve a BRUV unit as compared to conducting a UVC transect (T. Langlois pers. comm.), which may make BRUVs less costly than UVCs. Finally, BRUV provides a permanent record that can be repeatedly examined to ensure that fishes are correctly identified. However, there is a well-known bottleneck in the processing of BRUV tapes (Willis - 39 -

et al. 2000, Cappo et al. 2003, Stobart et al. 2007). We computed that ~2 UVC transects can be conducted, including field and lab time, for every single BRUV unit deployed. It is important to note that this is probably an overestimate of the time to deploy a BRUV unit because we were working with only 2 frames; use of additional frames will reduce this time considerably. In our research, the ratio of time to deploy a BRUV vs. a UVC means that many more BRUVs than UVCs are required to record a certain percentage of predicted species richness, with the result that UVCs may cost less in terms of personnel hours than BRUVs.

CONCLUSION Our comparison between 2 methods commonly used to survey subtidal fishes demonstrates that the method chosen to collect data will influence estimates of abundance. Using UVC, we obtained higher measures of species and family richness, diversity as measured by the Shannon Index, and recorded more individuals than using BRUV. We attribute the majority of these differences to the better ability of divers to search complex habitats as compared to a stationary camera. This is reflected in the significantly higher relative abundance and number of species of Type I recorded using UVC (Fig. 7a,b). In addition, it appears that BRUV underestimates the density of herbivorous and territorial species (Fig. 8a,b). In contrast, BRUV recorded higher species richness and abundance of mobile predators (Fig. 8c). Using BRUV we also recorded a more taxonomically diverse sample (Fig. 6a), though this difference was eliminated when taxonomic evenness was taken into account (Fig. 6b). These results suggest that studies wishing to catalogue diversity would do best to use a variety of methods, whereas those which target only a few select species would do best to select the method and size of sampling area which best suits those species. We are not alone in these recommendations: Many other studies have suggested that a combination of techniques may be necessary to survey an entire fish assemblage (Sale & Douglas 1981, Sale & Sharp 1983, Lincoln Smith 1988). However, in situations in which financial or time constraints limit researchers to only a single method, our results suggest that UVC will likely be the better option.

LITERATURE CITED Boulinier T, Nichols JD, Sauer JR, Hines JE, Pollock KH (1998) Estimating species richness: The importance of heterogeneity in species detectability. Ecology 79:1018-1028

- 40 -

Brock VE (1954) A preliminary report on a method of estimating reef fish populations. Journal of Wildlife Management 18:297-308 Cappo M (2006) Illustrated Guide to Assembly, Deployment and Retrieval of BRUVS. Australian Institute of Marine Science, p 34 Cappo M, Harvey E, Malcolm H, Speare P (2003) Potential of video techniques to monitor diversty, abundance and size of fish in studies of marine protected areas. In: Beuner JP, Grant A, Smith DC, Mahon D (eds) Aquatic Protected Areas: What works best and how do we know? Proceedings of the World Congress on Aquatic Protected Areas, Cairns, Australia; August 2002, p 455-464 Cappo M, Harvey E, Shortis M (2007) Counting and measuring fish with baited video techniques - an overview. In: Lyle JM, Furlani DM, Buxton CD (eds) Cutting-edge technologies in fish and fisheries science Australian Society for Fish Biology Workshop Proceedings. Australian Society for Fish Biology, Hobart, Tasmania, August 2006, p 101- 114 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. Journal of Experimental Marine Biology and Ecology 302:123-152 Castro ALF, Rosa RS (2005) Use of natural marks on population estimates of the nurse shark,Ginglymostoma cirratum, at Atol das Rocas Biological Reserve, Brazil. Environmental Biology of Fishes 72:213-221 Clarke KR, Warwick RM (1998) A taxonomic distinctness index and its statistical properties. Journal of Applied Ecology 35:523-531 Clarke KR, Warwick RM (2001a) Change in Marine Communities: An Approach to Statistical Analysis and Interpretation. Primer-E Ltd., Plymouth, U.K. Clarke KR, Warwick RM (2001b) A further biodiversity index applicable to species lists: variation in taxonomic distinctness. Marine Ecology-Progress Series 216:265-278 Colwell RK (2006) EstimateS: Statistical estimation of species richness and shared species from samples. Version 8.0. User's Guide and application published at: http://purl.oclc.org/estimates Edgar GJ (2005) Australian Marine Life: The Plants and Animals of Temperate Waters, Reed New Holland Publishers Pty Ltd, Sydney, Australia Edgar GJ, Barrett NS, Morton AJ (2004) Biases associated with the use of underwater visual census techniques to quantify the density and size-structure of fish populations. Journal of Experimental Marine Biology and Ecology 308:269-290 Eschmeyer WN, Fricke R (2009) Catalog of Fishes electronic version (9 September 2009). http://research.calacademy.org/ichthyology/catalog/fishcatmain.asp Farnsworth KD, Thygesen UH, Ditlevsen S, King NJ (2007) How to estimate scavenger fish abundance using baited camera data. Marine Ecology-Progress Series 350:223-234 Fowler AJ (1987) The development of sampling strategies for population studies of coral reef fishes - a case study. Coral Reefs 6:49-58 Friedlander AM, DeMartini EE (2002) Contrasts in density, size, and biomass of reef fishes between the northwestern and the main Hawaiian islands: the effects of fishing down apex predators. Marine Ecology Progress Series 230:253-264 Gomon MF, Bray D, Kuiter R (2008) Fishes of Australia's Southern Coast, Reed New Holland, Sydney Harvey E, Fletcher D, Shortis MR, Kendrick GA (2004) A comparison of underwater visual distance estimates made by scuba divers and a stereo-video system: implications for underwater visual census of reef fish abundance. Marine and Freshwater Research 55:573- 580 Harvey ES, Cappo M, Butler JJ, Hall N, Kendrick GA (2007) Bait attraction affects the performance of remote underwater video stations in assessment of demersal fish community structure. Marine Ecology-Progress Series 350:245-254

- 41 -

Heagney EC, Lynch TP, Babcock RC, Suthers IM (2007) Pelagic fish assemblages assessed using mid-water baited video: standardising fish counts using bait plume size. Marine Ecology-Progress Series 350:255-266 Hough D, Mahon G (1994) Biophysical classification of Victoria's marine waters. In: Muldoon J (ed) Towards a Marine Regionalisation for Australia. Great Barrier Reef Marine Park Authority, Sydney Kuiter RH (2000) Coastal Fishes of South-eastern Australia, Gary Allen Pty Ltd, Sydney Kulbicki M (1998) How the acquired behaviour of commercial reef fishes may influence the results obtained from visual censuses. Journal of Experimental Marine Biology and Ecology 222:11-30 Lincoln Smith MP (1988) Effects of observer swimming speed on sample counts of temperate rocky reef fish assemblages. Marine Ecology-Progress Series 43:223-231 Lincoln Smith MP (1989) Improving multispecies rocky reef fish censuses by counting different groups of species using different procedures. Environmental Biology of Fishes 26:29-37 MacNeil MA, Graham NAJ, Conroy MJ, Fonnesbeck CJ, Polunin NVC, Rushton SP, Chabanet P, McClanahan TR (2008a) Detection heterogeneity in underwater visual-census data. Journal of Fish Biology 73:1748-1763 MacNeil MA, Tyler EHM, Fonnesbeck CJ, Rushton SP, Polunin NVC, Conroy MJ (2008b) Accounting for detectability in reef-fish biodiversity estimates. Marine Ecology-Progress Series 367:249-260 Miller RJ (1975) Density of commercial spider crab, Chionoecetes opilio, and calibration of effective area fished per trap using bottom photography. Journal of the Fisheries Research Board of Canada 32:761-768 Patterson HM, Lindsay M, Swearer SE (2007) Use of sonar transects to improve efficiency and reduce potential bias in visual surveys of reef fishes. Environmental Biology of Fishes 78:291-297 Powter DM, Gladstone W (2008) The reproductive biology and ecology of the Port Jackson shark Heterodontus portusjacksoni in the coastal waters of eastern Australia. Journal of Fish Biology 72:2615-2633 Priede IG, Merrett NR (1996) Estimation of abundance of abyssal demersal fishes; a comparison of data from trawls and baited camera. Journal of Fish Biology 49:207-216 Raaijmakers JGW (1987) Statistical analysis of the Michaelis-Menten equation. Biometrics 43:793-803 Robbins WD, Hisano M, Connolly SR, Choat JH (2006) Ongoing Collapse of Coral-Reef Shark Populations. Current Biology 16:2314-2319 Sainte-Marie B, Hargrave BT (1987) Estimation of scavenger abundance and distance of attraction to bait. Marine Biology 94:431-443 Sale PF, Douglas WA (1981) Precision and accuracy of visual census technique for fish assemblages on coral patch reefs. Environmental Biology of Fishes 6:333-339 Sale PF, Sharp BJ (1983) Correction for bias in visual transect censuses of coral reef fishes. Coral Reefs 2:37-42 Samoilys MA, Carlos G (2000) Determining methods of underwater visual census for estimating the abundance of coral reef fishes. Environmental Biology of Fishes 57:289- 304 Sherrod PH (2008) Nonlinear Regression Analysis Program (NLREG), Version 6.4, Nashville, TN, USA 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 assemblages: Methodological evaluation. Journal of Experimental Marine Biology and Ecology 345:158-174 Watson DL, Harvey ES (2007) Behaviour of temperate and sub-tropical reef fishes towards a stationary SCUBA diver. Marine and Freshwater Behaviour and Physiology 40:85-103

- 42 -

Watson DL, Harvey ES, Anderson MJ, Kendrick GA (2005) A comparison of temperate reef fish assemblages recorded by three underwater stereo-video techniques. Marine Biology 148:415-425 Watson DL, Harvey ES, Kendrick GA, Nardi K, Anderson MJ (2007) Protection from fishing alters the species composition of fish assemblages in a temperate-tropical transition zone. Marine Biology 152:1197-1206 Willis TJ (2001) Visual census methods underestimate density and diversity of cryptic reef fishes. Journal of Fish Biology 59:1408-1411 Willis TJ, Babcock RC (2000) A baited underwater video system for the determination of relative density of carnivorous reef fish. Marine and Freshwater Research 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. Marine Ecology-Progress Series 198:249-260 Yau C, Collins MA, Bagley PM, Everson I, Nolan CP, Priede IG (2001) Estimating the abundance of Patagonian toothfish Dissostichus eleginoides using baited cameras: a preliminary study. Fisheries Research 51:403-412

- 43 -

CHAPTER TWO

Body size and spatial grain influence fish-habitat associations in a temperate reef fish assemblage

ABSTRACT Understanding the scale at which species interact with their environment is essential for management and conservation. The importance of considering spatial extent, i.e., the area over which sampling is conducted, is well recognized, while the influence of spatial grain, i.e., the size of the sampling unit, has received less attention. In this research, we investigate the spatial grain at which temperate marine fishes from southern Australia associate with nearshore rocky reef habitat, the influence that body size has on animal-habitat relationships, and how body size affects the habitat characteristics with which species associate. We found that small-bodied species were infrequently associated with simple habitats, which we suggest is related to size- specific predation risk and swimming ability. We also found that correlations between all species and habitat were strongest in coarse grain models, and that large-bodied species only exhibited a strong association with habitat when it was measured at a coarse grain. These results suggest that fish-habitat associations should be investigated at a coarser grain than is typical, and underscore the importance of considering the scale at which species relate to their environment when selecting the spatial grain at which to quantify animal-habitat associations. That species respond to habitat heterogeneity at a larger scale than was previously thought, has implications for the management and conservation of marine resources.

INTRODUCTION Identifying the underlying causes of species distributions is a fundamental goal of ecology. At local scales, species’ distributions are often governed by habitat characteristics that influence the availability of resources (e.g., shelter, food, mates) (Grinnell 1917, Underwood et al. 2004). Ecological processes utilizing these resources take place on a variety of spatial scales (Wiens 1989, Sale 1998) – for example, foraging occurs at a smaller spatial scale than dispersal (Boyce 2006). How species respond to habitat is dependent not only upon the scale at which their response

- 44 - is measured (e.g., Wellnitz et al. 2001, Underwood et al. 2004), but also upon the species themselves (e.g., Boyce 2006, Meyer and Thuiller 2006). For example, species vary in their diet and capacity for dispersal. The conclusions reached by studies aiming to elucidate mechanisms driving species’ distributions will therefore depend upon the scale at which the study is conducted and the taxon-specific ecological processes acting at that scale (Wiens 1989, Sale 1998, Underwood et al. 2004). Consequently, adopting a multi-scale approach often strengthens associations between animals and habitat (e.g., Eagle et al. 2001, Meyer and Thuiller 2006, Grober-Dunsmore et al. 2008, Boscolo and Metzger 2009).

The study of animal-habitat associations must take into account both large- and small- scale environmental variation, and should consider the scale at which the study taxa relate to their environment. At a large scale, the behaviour of many species, including interactions with habitat, can vary geographically (Foster 1999). Many studies have explicitly considered the scales at which animals respond to their environment by varying spatial extent, i.e., the area over which sampling is conducted (Wiens 1989). The approach such studies frequently employ is to measure how species interact with environmental variables as a function of the distance at which observations are made across a landscape (e.g., Wellnitz et al. 2001, Boscolo and Metzger 2009, Feist et al. 2010). At a small scale, it is important to consider the spatial grain of a study, which Wiens (1989) defined as the finest level of spatial resolution possible within a data set, i.e., the sampling unit. If the grain of a study is too small, it will not include the full range of habitat heterogeneity and will consequently underestimate the response of species to their environment (Boyce 2006). Therefore, a study’s perception of how a species responds to environmental heterogeneity will be influenced by its choice of both spatial extent and grain (Wiens 1989).

The choice of spatial extent and grain should be appropriate to the study taxa – i.e., the selection should consider the scale at which species relate to their environment (Wiens 1989, Sale 1998, Meyer and Thuiller 2006). The ‘functional grain’ is the smallest scale at which an organism responds to habitat heterogeneity (Baguette and Van Dyck 2007), and the coarseness of this grain is ultimately a function of body size (Brown 1984, Holling 1992), as body size affects the scale at which an organism perceives its environment ( Wiens 1989, Baguette and Van Dyck 2007).

- 45 -

Figure 1: The number of habitat, BRUV and UVC samples per location and the dates during which sampling occurred.

There are good reasons to expect that species of different body sizes will therefore interact with the environment at different spatial scales. Body size has been shown to influence the need for shelter (Anderson et al. 1989) and space (Biedermann 2003), acquisition of resources (Ritchie and Olff 1999), home range size (Holling 1992), and extent of movement (Roland and Taylor 1997). Despite the potential for body size to influence animal-habitat associations, little research has explicitly explored how the scale at which habitat is measured affects differently-sized species. What research has been conducted has found that larger species are more strongly correlated with habitat when it is measured at a larger scale (Roland and Taylor 1997, Holland et al. 2005, Murakami et al. 2008).

- 46 -

Marine reef systems are inherently multi-scalar and investigation into associations between reef fishes and habitat must take this into account (Sale 1998). For example, in the U.S. Virgin Islands, the scale at which coral reef fishes respond to their environment varies, with fine-scale measures accurately predicting the occurrence of some species while others respond to landscape-scale predictors (Grober-Dunsmore et al. 2008). Research into fish-habitat associations has identified several consistent patterns. First, there is abundant evidence to suggest that fish species richness, diversity and density are positively associated with habitat complexity (e.g., Choat and Ayling 1987, Connell and Jones 1991, Willis and Anderson 2003). Second, in temperate marine systems the fish assemblage associated with kelp-dominated areas is often different to the one associated with urchin barrens (e.g., Choat and Ayling 1987, Willis and Anderson 2003, Anderson and Millar 2004). And finally, depth appears to play a role in determining the distribution of many species (e.g., Cappo et al. 2007, Pérez-Matus et al. 2007, Chatfield et al. 2010).

While there have been many investigations into how fish assemblages covary with habitat, investigations into fish-habitat associations as a function of body size have focused on ontogenetic changes in habitat use (e.g., Anderson et al. 1989, Dahlgren and Eggleston 2000), feeding preferences (e.g., Choat and Ayling 1987) or dispersal capacity as it relates to geographic range size (e.g., Reaka 1980). To the best of our knowledge, no research has investigated how the spatial grain of a study and body size influence fish-habitat associations, despite the many reasons to expect such relationships to exist. For example, large-bodied fishes appear to have different relationships to habitat than small-bodied species (Choat and Ayling 1987, Anderson et al. 1989), differences that could be driven by size-specific predation risk and/or swimming abilities.

The positive relationship between fish density and habitat complexity has been attributed to a reduction in exposure both to predation (Beukers and Jones 1998, Shoji et al. 2007) and water movement from waves and currents (Fulton et al. 2001, Johansen et al. 2008). Because both predation risk and swimming ability are dependent on body size, differently sized species may have different requirements for habitat complexity. Complex habitats reduce predation risk (Beukers and Jones 1998, Shoji et al. 2007) and so may be important particularly for small-bodied species.

- 47 -

Interestingly, exposure to predation in areas with no refuge has been shown to reduce the maximum body length of pelagic crustaceans (Slusarczyk and Pinel-Alloul 2010).

Water movement can have a large effect on a species’ morphology, and has been implicated as a driving force in the evolution of hydrodynamic body shape in fish (e.g., Sagnes et al. 2000). In the temperate marine environment, body size is more strongly correlated with swimming ability than fin shape (Fulton and Bellwood 2004, Denny 2005) and superior swimmers can better deal with surge (Johansen et al. 2008). Refuging behaviour can mediate poor swimming ability (Johansen et al. 2008), but only where refuges exist. As the number of refuges is integrally linked to the degree of habitat complexity, we might expect to only find smaller-bodied species in areas of high surge, i.e., shallow waters, if the habitat is complex.

In this research, we investigated associations between nearshore rocky reef fishes and habitat in south-eastern Australia. In addition to quantifying fish-habitat associations, we explored whether spatial grain and/or body size influences relationships between fish and habitat. Specifically, we predicted that including coarser grains in models would strengthen the relationships between fish and habitat. Secondly, we expected the relationship between fish and spatial grain to be dependent on body size, with the specific prediction that smaller-bodied species would be more strongly associated with habitat at smaller scales than larger-bodied species. Finally, we predicted that smaller-bodied species would be associated with areas of high structural complexity that would facilitate their avoidance of predation and/or wave exposure.

MATERIALS AND METHODS Data were collected at three locations along the open coast of Victoria in south- eastern Australia between December 2007 and June 2008 (Fig. 1). Surveys of the fish assemblage were conducted using two methods: underwater visual census (UVC) and baited remote underwater video (BRUV). At each location, approximately 1 habitat survey was conducted for each BRUV unit deployed, and approximately 3 habitat transects conducted and 3 BRUV units deployed for each UVC. The actual number of surveys of each type is listed in Fig. 1. All surveys were conducted between 3.7 and 28 m depth. Whenever conditions allowed it, habitat surveys were conducted on the

- 48 - same day as fish surveys. At most, 20 days elapsed between a habitat survey and a fish survey, with a mean of 4 days (± 5 SD) elapsed.

Fish surveys BRUV units were comprised of a Sony HC-series Handycam in an underwater housing mounted onto an aluminium frame. Bait, consisting of 400 g of crushed pilchards, was placed in a mesh pouch that was suspended from a PVC pipe 1.5 m from the camera lens. The unit was deployed and retrieved remotely onto or next to rocky reef, and was immersed for approximately 60 minutes. After deployment, the boat motored away from the area. The footage was later viewed on a computer monitor and a measure of relative abundance, MaxN, enumerated. MaxN is the maximum number of individuals of each species observed in a single frame or one- second interval on the tape. All tapes were viewed by one observer and, because the observer’s ability to identify fishes improved over time, the first 12 tapes to be viewed were watched a second time.

UVC transects were conducted by an observer and buddy on SCUBA. The divers entered the water in the location in which a BRUV had been deployed and swam slowly in a pre-determined direction while the observer counted and identified fish in a strip 5 m wide. The direction of each transect was chosen prior to the dive based on current velocity, presence of rocky reef, and the ability of the boat to safely retrieve divers. Transect duration was limited by no-decompression limits in dive tables and air supply. The observer looked in the water column as well as through kelp and under overhangs for fishes, while the second diver, present for safety purposes, remained behind the observer. Transect length was computed using the GPS coordinates of diver entrance and egress points. Survey methodology is more fully described in Colton and Swearer (2010).

Taxonomic nomenclature follows Eschmeyer and Fricke (2009).

Habitat Habitat was measured by divers using SCUBA along two 50 m transects radiating out haphazardly from the location in which a BRUV had been previously deployed. At 1 m intervals along each transect, a diver measured the depth of the water column at the

- 49 -

Table 1: Physical and biological habitat categories and descriptions. Category Sub-category Description Rugosity n/a Ratio of contour distance to straight line distance Depth n/a Mean depth (m) Substrate Abiotic with no animal, plant or macroalgae Pave Pavement: solid rock; no edges; ≥ ⅓ of photo Boul Boulder: ≥3 edges visible; rock must be ≤ ½ size picture Rubb Rubble: grains visible; mixture of sand and rock Sand Including shell pieces but not rock; small grain size Cave Overhang or hole in substrate Uk-subs Unidentifiable substrate; no animal, plant or algal growth Structural Tall+long Animal and algae/plant ≥10 cm tall/long attributes Short+tuft Animal and algae/plant < 10 cm tall/long Encr Encrusting animal + encrusting algae Remn+Drift Algae/plant matter no longer attached to substrate + obviously non-living animal Uk_length Animal, plant/algae of unidentifiable height or length Biotic_total Total area covered by Animal and MA MA Total area covered by macroalgae and plant Long Algae/plant frond length ≥10 cm Tuft Algae/plant frond length < 10 cm Encr Encrusting algae Drift Algae/plant matter no longer attached to substrate Uk-le Algae/plant of unidentifiable height Animal Total area covered by Animal Short Animal height ≥10 cm Tall Animal height < 10 cm Enc Encrusting animal Remn Obviously non-living animal (e.g., part of a shell) Unk Unknown: unidentifiable MA Uk_ma Unknown macroalgae Er Ecklonia radiata Carp Carpoglossum confluens Durv Durvillaea potatorum Cyst Cystoseiraceae Seir Seirococcaceae Sarg Sargassaceae C-alg Corallinaceae Grac Gracilariaceae Amph Amphibolis spp. Posi Posidonia spp. Chlor Sum of area covered by all Chlorophyta Phae Sum of area covered by all Phaeophyta Rhod Sum of area covered by all Rhodophyta Seag Sum of area covered by all seagrasses Animal Spon Porifera Anth Anthozoa Hydr Hydrozoa Shel Shelled animals, including Cirripedia & Mollusca Bryo Bryozoa Star Asteroidea & Crinoidea Urch Echinoidea Asci (Ascidiacea Cora Scleractinia & Alyonacea Uk-a Unknown Animalia

- 50 - top of the substrate using a depth gauge, timing measurements with swell to ensure consistency. A second diver took photographs at 5 m intervals along the same transect tape, placing a 1 m piece of white PVC pipe marked with 10 cm bands on the substrate approximately perpendicular to the transect tape, floating above the stick until its entirety could be seen in the viewfinder, and taking a picture using a point- and-shoot camera mounted in an underwater housing.

The depth measurements were used to compute a contour distance (sensu McClanahan and Shafir 1990) by solving Pythagoras’ theorem for the hypotenuse, with the legs of the triangle equal to the 1 m distance separating depth measurements, and the difference between two adjacent depth measurements. Rugosity was computed as the ratio of contour distance to straight-line (transect) distance.

Habitat photographs were downloaded onto a computer and viewed using the program CPCe (Kohler and Gill 2006). Using the Calibrate routine and the metre-stick or visible portion thereof, the length and width of each photograph were measured. For photographs in which the view of the metre-stick was occluded by kelp, the mean length and width of other photographs from that transect were used to approximate the area of the photograph. Onto each photo a grid of 5 x 10 cells was laid and a point randomly assigned within each cell, again using CPCe. The substrate under each point was identified and classified based first on whether it was abiotic or biotic, second on its physical attributes (i.e., structure), and third on the lowest level of taxonomic classification possible (Table 1). The physical attributes described the type of substrate present, if any, and the height that any biological matter extended above the benthos. Taxonomic resolution was rarely possible beyond class for animals, family for macroalgae and genus for plants. The habitat under the random point was assumed to represent the entire contents of the cell in which it fell. The dimensions of each cell were computed based on the total length and width of each photograph, and the number of points of each substrate type were multiplied by the area of each cell and summed across the two 50 m transects to provide the total area covered by each habitat type.

- 51 -

Data analyses Because we measured abundance using two methods and because the habitat data had to be paired to the abundance data in two different ways, we conducted two sets of analyses exploring relationships between habitat data and (1) BRUV, and (2) UVC. BRUV, habitat and UVC were matched based on the physical proximity of replicate surveys using GPS coordinates. BRUV deployments were easily paired with habitat transects both because the habitat transects originated from the site of a BRUV deployment, and because there was generally a 1:1 ratio of BRUV deployments to habitat transects. Data from up to three habitat transects were combined in order to be paired with data from a single UVC transect. The amount of area covered by each habitat category (Table 1) was summed across the three transects, and both depth and rugosity were averaged. The abundance data collected by BRUV and UVC (MaxN and density, respectively) were fourth-root transformed to correct for skew, which typify abundance data, and analyses conducted on Bray-Curtis resemblance matrices.

To explore the scale at which fish and habitat were most closely associated, we cumulatively aggregated habitat data to the ratios:

1. 1 BRUV : 1 Habitat = 100 m of habitat data 2. 3 BRUV : 1 UVC : 3 Habitat = 300 m of habitat data 3. 6 BRUV : 2 UVC : 6 Habitat = 600 m of habitat data 4. 12 BRUV : 4 UVC : 12 Habitat = 1200 m of habitat data 5. 24 BRUV : 8 UVC : 24 Habitat = 2400 m of habitat data

To ensure a fully balanced design in which 8 UVC transects, 24 BRUV deployments and 24 habitat transects were completed at each location, mean data, computed using all replicates at a location, were used in the cases in which replicates were missing (see Fig. 1). At each level of aggregation, we summed MaxN across all transects, and summed all habitat variables except for depth and rugosity, which were averaged. For the UVC data, we computed a density for each species at each level of aggregation as the sum of the number of individuals observed in n transects divided by the total area surveyed in n transects. It is important to note that we used summed data at each level of aggregation, as averaging would result in the loss of variation, potentially resulting

- 52 - in stronger, but spurious, relationships between fish and habitat with increasing spatial grain.

Analysing these data using a multivariate multiple regression approach was found to be problematic due to high covariance among the habitat variables. We exploited this spatial covariance by identifying habitat types using a group-average clustering method applied with the Similarity Profile (SIMPROF) routine in Primer-E (Clarke and Warwick 2001), with significance assessed using 999 permutations. The significant clusters were then used as factors in a non-parametric multivariate analysis of variance with significance assessed by permutations (PERMANOVA) to test whether the abundance data were significantly associated with habitat clusters, again using PRIMER-E. Principal components ordinations (PCOs) were used to visualize the results and to explore species-specific relationships to habitat variables.

We explored how body size influenced the scale at which fish associate with habitat by exploring interactions between species body size and aggregation level. Species were placed into one of four arbitrarily selected body size classes based on their reported maximum total length ( Kuiter 2000, Gomon et al. 2008): (TL1) 1 – 30cm, nBRUV = 13 species, nUVC = 21 species; (TL2) 31 – 60 cm, nBRUV =21 species, nUVC =

20 species; (TL3) 61 – 100 cm, nBRUV =11 species, nUVC = 7 species; and (TL4) > 100 cm , nBRUV =10 species, nUVC = 5 species. We examined changes in the strength of Spearman correlation (|ρ|) between fish and habitat for species in each of these four categories by examining correlations between abundance, measured using BRUV or UVC, and the first two PCO axes. We used |ρ| as some relationships between fish abundance and PCO axes were negative. Tests were conducted using SPSS (v. 16.0) on square-root transformed |ρ| using a repeated measures MANOVA with two factors: total length (random) and level of aggregation (fixed). Multivariate ANOVAs were selected as the data violated the assumption of sphericity (Mauchley’s W p < 0.05).Using the BRUV data, we explored four levels of aggregation (100 m, 200 m, 600 m and 1200 m) as our analyses showed no significant clusters at 2400 m. With the UVC data, we explored three levels of aggregation (300 m, 600 m and 1200 m) as a single UVC transect covered, on average, 300 m of habitat.

- 53 -

0.9 35 0.8 30 0.7 25 0.6 0.5 20 0.4 15 0.3 10 0.2 0.1 5

Spearman correlation (rho) correlation Spearman 0 0 Number of significant of Number clusters significant 100m 300m 600m 1200m 2400m

Level of aggregation Figure 2: Variation in the strength of Spearman correlation, |ρ|, and the number of significant clusters (solid line) as a function of the level of aggregation of habitat data for data collected by BRUV (grey) and UVC (white). PERMANOVA tests for all but the highest level of aggregation were significant at p<0.005; tests were not possible at the highest level of aggregation due to a lack of significant clusters.

RESULTS Using a coarser grain, i.e., aggregate habitat data, strengthened the association between fish and habitat. When 1200 m of habitat data were used, four significant clusters were identified in the habitat data and PERMANOVA tests exploring relationships between species and habitat were significant (Fig. 2), suggesting that 1200 m of habitat may be the appropriate scale at which to assess fish-habitat associations at the assemblage level in this study.

We examined PCOs at all levels of aggregation and found little difference in the habitat variables their axes represented, so we only report results based on 1200 m of habitat data. At this level, the first two PCO axes together explained 61.3% of the variation in habitat variables. The third PCO axis explained an additional 19.6%, but was comprised only of components represented in the first two axes. As such, it neither changed nor added to our interpretation of these data, and so is not discussed further. Vectors representing the strength of correlations between habitat variables and the PCO axes were overlaid on the PCO plots (Fig. 3a) to visualize the habitat characteristics responsible for the axes. PCO axis 1 was strongly associated with depth: shallow-water species, e.g., seagrasses, had low PCO1 scores while deeper- water species, e.g., sponges, had high PCO1 scores (Fig. 3a,b). The second axis was associated with changes from biotic to abiotic cover (e.g., from Phaeophyta (brown algae) to sand) and also with increasing structural complexity. Tall and long

- 54 - organisms had low PCO2 scores while encrusting and abiotic variables had high PCO2 scores (Fig. 3a,b).

Associations between species and habitat were explored using species-specific correlations with PCO axes (Fig. 3c,d, Table 2,3). Using a model that included 1200 m of habitat data, the BRUV data revealed several species to be associated with shallow and simple/abiotic habitat (Fig. 3b), including Trygonorrhina spp., Myliobatis australis, Cephaloscyllium laticeps, Pseudolabrus psittaculus, and Heterodontus portusjacksoni (Fig. 3c). In the UVC data, only four species, Meuschenia scaber, Pseudocaranx spp., Thamnaconus degeni and Trygonorrhina spp. were associated with these habitat characteristics (Fig. 3d). Several species in both the BRUV and UVC data were associated with deep simple habitats (Fig. 3b,c,d). In the BRUV data, several species were associated with shallow, complex habitats (Fig. 3b), whereas in the UVC data, only two species, Trachichthys australis and Meuschenia hippocrepis, were associated with this type of habitat. In both data sets, several species were associated with deeper, more complex habitats (Fig. 3b), including Scorpis lineolata and Pempheris multiradiata.

As the amount of habitat data included in the models increased, so did the number of species that were strongly correlated with habitat (Fig. 4). When the BRUV data were associated with 300 m of habitat, only one species had |ρ| > 0.75, whereas at 1200 m, 23 species had |ρ| > 0.75 (Table 2). When the UVC data were associated with 300 m of habitat data only one species had |ρ| > 0.75, whereas at 1200 m 16 species had |ρ| > 0.75 (Table 3). The strength of correlation also increased with increasing amounts of habitat data: when the BRUV and UVC data were associated with 300 m of habitat, the maximum value of |ρ| = 0.81, whereas at 1200 m, the maximum |ρ| = 0.94.

The number of large-bodied species (i.e., TL4) observed using BRUV increased with the inclusion of more habitat data. When 300 and 600 m of habitat data were used, only one TL4 species, Chrysophrys auratus, was strongly associated with habitat, but when 600 m of habitat data was used, 7 TL4 species were strongly associated with habitat. In the UVC data, no TL4 species were strongly associated with habitat at the lowest level of aggregation, one TL4 species, Trygonorrhina spp., associated with 600

- 55 -

Table 2: For data collected by BRUV, species that showed correlations ρ ≥ 0.60 with the first two PCO axes based on significant clusters by level of aggregation (m). TL = total length category (see ‘Methods’). Species are listed in order of decreasing correlation strength Note: 100 m is not shown because no species had |ρ| ≥ 0.60 at this level. 300 m 600 m 1200 m PCO Species TL ρ PCO Species TL ρ PCO Species TL ρ 1 Chrysophrys auratus 4 -0.67 1 Chrysophrys auratus 4 -0.67 1 Caesioperca rasor 1 0.94 2 Meuschenia galii 2 -0.76 Caesioperca rasor 1 0.62 Atypichthys strigatus 1 0.85 Meuschenia flavolineata 1 -0.64 Meuschenia hippocrepis 2 -0.60 Caesioperca lepidoptera 2 0.85 2 Cheilodactylus spectabilis 3 0.79 Conger verreauxi 4 0.85 Parma victoriae 1 -0.79 Dotolabrus aurantiacus

/Eupetrichthys angustipes 1 0.85 Latridopsis forsteri 3 0.72 Latris lineata 4 0.85 Eeyorius hutchinsi 1 0.70 Meuschenia australis 2 0.85 Haletta semifasciata 2 0.70 Notorynchus cepedianus 4 0.85 -

56 Meuschenia galii 2 -0.67 Pentaceropsis recurvirostris 3 0.85

- Neosebastes scorpaenoides 2 0.67 Chrysophrys auratus 4 -0.77

Paraplesiops meleagris 2 -0.65 Scorpis aequipinnis 3 -0.77 Pempheris multiradiata 1 -0.64 Scorpis lineolata 2 0.76 Sphyraena novaehollandiae 3 0.63 Cheilodactylus nigripes 2 0.75 Odax cyanomelas 2 -0.61 Meuschenia hippocrepis 2 -0.64 Enoplosus armatus 2 -0.60 Upeneichthys spp. 2 0.60 Aplodactylus arctidens 3 0.60 Dinolestes lewini 3 0.60 2 Cephaloscyllium laticeps 4 0.94 Odax cyanomelas 2 -0.89 Meuschenia galii 2 -0.85

- 56 -

Table 2 Continued 300 m 600 m 1200 m PCO Species TL ρ PCO Species TL ρ PCO Species TL ρ Heterodontus portusjacksoni 4 0.83 Notolabrus tetricus 2 -0.83 Latridopsis forsteri 3 0.81 Pseudophycis barbata 3 0.81 Trachinops caudimaculatus 1 -0.78 Dinolestes lewini 3 0.77 Parma victoriae 1 -0.77 Upeneichthys spp. 2 0.77 Enoplosus armatus 2 -0.71 Pseudolabrus psittaculus 2 0.71 Meuschenia flavolineata 1 -0.70 -

57 Eubalichthys mosaicus 2 -0.65

- Paraplesiops meleagris 2 -0.65

Pempheris multiradiata 1 -0.65 Myliobatis australis 4 0.60 Scorpis aequipinnis 3 -0.60

- 57 -

Table 3: For data collected by UVC, species that showed correlations ρ≥ 0.60 with the first two PCO axes based on significant cluster s by level of aggregation (m). TL = total length category (see ‘Methods’). Species are listed in order of decreasing correlation strength. 300 m 600 m 1200 m PCO Species TL ρ PCO Species TL ρ PCO Species TL ρ 1 Enoplosus armatus 2 0.75 1 Caesioperca rasor 1 0.92 1 Enoplosus armatus 2 0.94 Caesioperca rasor 1 0.65 Enoplosus armatus 2 0.84 Pictilabrus laticlavius 1 0.94 Upeneichthys spp. 2 0.60 Pictilabrus laticlavius 1 0.81 Caesioperca rasor 1 0.90 2 Meuschenia flavolineata 1 -0.81 Pempheris multiradiata 1 0.79 Acanthaluteres vittiger 2 0.89 Upeneichthys spp. 2 0.77 Eupetrichthys angustipes 1 0.88 Acanthaluteres vittiger 2 0.76 Siphonognathus beddomei 1 0.85 Siphonognathus beddomei 1 0.66 Cheilodactylus spectabilis 3 0.84 Parma microlepis 1 0.65 Upeneichthys spp. 2 0.83 Trachichthys australis 1 -0.65 Parequula melbournensis 1 0.78 Atypichthys strigatus 1 0.63 Parma microlepis 1 0.78 -

58 Eupetrichthys angustipes 1 0.62 Parma victoriae 1 0.77

Notolabrus tetricus 2 0.62 Pempheris multiradiata 1 0.77 Parma victoriae 1 0.61 Atypichthys strigatus 1 0.76 2 Meuschenia flavolineata 1 -0.75 Aracana aurita 1 0.70 Trygonorrhina spp. 4 0.72 Dactylophora nigricans 4 0.70 Trachinops caudimaculatus 1 -0.65 Meuschenia australis 2 0.70 Odax acroptilus 1 0.70 Pseudolabrus psittaculus 2 0.70 Scorpis lineolata 2 0.68 Aracana ornata 1 0.65 Trachichthys australis 1 -0.65 Trachurus novaezelandiae 2 0.65

- 58 -

Table 3 Continued 300 m 600 m 1200 m PCO Species TL ρ PCO Species TL ρ PCO Species TL ρ Meuschenia hippocrepis 2 -0.60 Notolabrus tetricus 2 0.60 Odax cyanomelas 2 0.60 Meuschenia flavolineata 1 -0.82 Dotolabrus aurantiacus 1 -0.78 Odax cyanomelas 2 0.77 Odax acroptilus 1 0.70 Meuschenia scaber 2 0.65 Notolabrus gymnogenis 2 -0.65 Thamnaconus degeni 1 0.65

- Trachinops caudimaculatus 1 -0.65

59 Aracana aurita 1 0.64

- Pseudolabrus psittaculus 2 0.64 Meuschenia freycineti 2 0.60 Upeneichthys spp. 2 0.60

- 59 -

m of habitat data, and one TL4 species, Dactylophora nigricans, associated with 1200 m of habitat data.

In addition to body size affecting the spatial grain at which species responded most strongly to habitat, there also appeared to be some size-specific responses to habitat variables. Notably, few small-bodied species (i.e., TL1) were positively associated with shallow, simple habitat: Parequula melbournensis in the BRUV data (Fig. 3c), and Thamnaconus degeni, Odax acroptilus and Aracana aurita in the UVC data (Fig. 3d). Of these, only T. degeni was observed in shallow environments.

Repeated measures MANOVAs on BRUV and UVC data revealed a significant effect of aggregation level for each PCO axis, no significant effect of total length, and only one significant interaction between aggregation and total length for UVC data on PCO1 (Table 4). However, plots of |ρ| as a function of aggregation level for each of the four body size classes revealed interesting patterns (Fig. 5). The BRUV data generally showed an increase in correlation strength as a function of increasing amounts of habitat data. For the lowest level of aggregation on PCO1, the two smaller body size classes (i.e., TL1 and TL2) had higher correlations than the two larger classes (i.e., TL3 and TL4) (Fig. 5a). For PCO2, increases in correlation strength for the largest species did not occur until the highest level of aggregation, though correlation strength increased for smaller species on PCO2 when 300 m of habitat data were included (Fig. 5b). The UVC data showed similar patterns. On PCO1, only the smallest species had |ρ| > 0.5 at the lowest level of aggregation and showed no increase in correlation strength when the amount of habitat data increased from 600 m to 1200 m (Fig. 5c). In contrast, species in the third size class did not show an increase in correlation strength until 1200 m of habitat data were included. For PCO2, the smallest three size classes showed increasing correlation strength with increasing aggregation, while the correlation strength for the largest size class was highest when 600 m of habitat data were used (Fig. 5d). Interestingly, the largest sized species showed little improvement in correlation strength with increases in aggregation level.

DISCUSSION In this research, increasing the coarseness of habitat grain strengthened fish-habitat associations. Many studies investigating relationships between fish and habitat

- 60 -

Figure 3 (This and following page): Principal components ordinations (PCOs) based on significant clusters from 1200 m of habitat data. Vectors representing Spearman correlations with |ρ| > 0.6 between PCO axes and (a) habitat variables, (c) MaxN, and (d) density measured using UVC. (b) Shows photographic examples of habitat types. Full species’ names in are provided in Tables 3 and 4.

- 61 -

Figure 3 Continued

- 62 -

Table 4: Results of a repeated measures MANOVA testing for the effects of spatial grain (level of aggregation) and body size on the strength of the relationship (|ρ|) between fish abundance and habitat. The F-statistic is Pillai’s Trace, Agg = level of aggregation, and TL = total length category. Significant p-values are in bold. BRUV UVC Factor PCO# F df p F df p Agg 1 8.81 3,49 <0.00005 16.13 2,50 <0.00005 TL 0.96 3 0.42 1.83 3 0.15 Agg * TL 0.63 9,153 0.77 3.19 6,102 0.007 Agg 2 32.02 3,49 <0.00005 5.65 2,50 0.006 TL 1.87 3 0.15 0.40 3 0.75 Agg * TL 0.76 9,153 0.66 1.37 6,102 0.23

measure habitat along 25 or 50 m transects (e.g., Choat and Ayling 1987, Anderson and Millar 2004). Just because this is a convenient unit of measurement does not mean that species will respond at this spatial grain (Wiens 1989, Sale 1998). In fact, we found that models that included more habitat data revealed stronger relationships between fish and habitat (Fig. 2,4) and that, in this study, including 1200 m of habitat data appeared to best model fish-habitat associations. In New Zealand, Anderson and Millar (2004) noted that it was difficult to predict some species’ distributions using habitat data measured at a small scale. Similarly, Feist et al. (2010) found that predictor variables were most strongly correlated with salmon redd density when they were summarized over a catchment rather than at the scale of a stream. Studies such as these underscore the importance of varying the extent of a study; our results suggests that it is also important to consider the spatial grain at which habitat is measured. In a meta-analysis of animal resource selection functions, Meyer and Thuiller (2006) reported that multi-grain models were more predictive of species’ distributions across their range than single-grain models. In this research, we found that increasing the size of the grain strengthened relationships between fish and habitat. It is possible that these associations would have been stronger had more habitat data been included, but because of low power, we were unable to conduct statistical tests at higher levels of aggregation.

Increasing grain coarseness may not improve fish-habitat relationships for all species in all systems because as the coarseness of the grain is increased, the ability to detect fine-scale patterns diminishes (Wiens 1989). Neither UVC nor BRUV are especially good at surveying small, cryptic species (Willis 2001, Colton and Swearer 2010), and

- 63 -

these species are most likely to respond to habitat at a fine grain. A study of associations between cryptic fishes and habitat in New Zealand found that abundance and composition were strongly influenced by reef complexity when habitat was measured in 9 m2 plots (Willis and Anderson 2003). Thus, our conclusion that species respond to habitat at a coarser spatial grain may be related to the types of fish we surveyed. Indeed, measuring associations between fish and habitat at the assemblage level is likely to require a multi-grain approach (Meyer and Thuiller 2006) that incorporates scales of meters to kilometres as different components of the assemblage will respond to habitat variability across this range of scales (Fig. 4).

We found a size-specific response of species to the coarseness of the spatial grain. When few habitat data were included in a model, only smaller-bodied species (i.e., TL1 and TL2) tended to be strongly associated with habitat, whereas larger-bodied species (i.e., TL3 and TL4) tended to be associated with habitat only at larger spatial grains (Fig. 5). Similarly, larger species of parasitic flies (Roland and Taylor 1997), longhorned beetles (Holland et al. 2005), and forest birds (Murakami et al. 2008) responded to forest cover when it was measured at larger scales. That larger-bodied species tend to experience the environment at larger spatial scales is reflected in the fact that they tend to have larger home range sizes (Jenkins 1981, Swihart et al. 1988). In addition, this is also likely driven by the fact that the minimum scale at which an animal perceives its environment is constrained by its body size (Holling 1992). Large species are incapable of experiencing the environment at as fine a scale as small species, which is related to the inherently fractal nature of the environment (Brown 1984).

In the marine environment, both predation risk (Magnhagen and Borcherding 2008) and swimming ability ( Fulton and Bellwood 2004, Denny 2005) are related to body size. As habitat complexity can reduce predation risk (Beukers and Jones 1998, Shoji et al. 2007) and exposure to flow (Johansen et al. 2008), we predicted that smaller- bodied species would not only respond to habitat at a smaller scale, but would also be more reliant on shelter than larger-bodied species. We found that smaller-bodied species were infrequently associated with simple habitats, and that those species which were associated with simple habitats generally occurred in deeper areas where, presumably, wave exposure was reduced. Similarly, Guidetti and Bussotti (2000)

- 64 -

100 |

rho 80

BRUV 40 UVC

20 Percent of species with max | max with species of Percent 0 100 m 300 m 600 m 1200 m

Level of habitat aggregation Figure 4: The percent of species that had their strongest association with habitat (i.e., |ρ|) at each level of habitat aggregation. found that small-bodied fishes were only found at more sheltered sites in the Mediterranean.

For many animals, acquiring food requires leaving the safety of a refuge, and so foraging becomes a compromise between resource acquisition and mortality risk (Werner 1992). In our research, the few species that were associated with shallow, simple habitats included several rays (e.g., Trygonorrhina spp. and Myliobatis australis), which feed on soft substrates and are morphologically well-suited to deal with surge. Upeneichthys spp. (TL2), which were associated with abiotic substrate in relatively deep areas in both BRUV and UVC data, feed by using their barbels to locate invertebrates buried in the sand. While their relatively small size would suggest that they would preferentially associate with structure in order to avoid predation and/or surge, their diet requires them to spend a large amount of time on sand. They were most frequently observed in small groups of ~ 6 individuals (pers. obs.), which may mitigate their predation risk, and their association with deeper areas may indicate an inability to deal with surge over sand in the shallows. Other smaller-bodied species that were observed in shallow, simple habitats include Meuschenia scaber and Thamnaconus degeni, which are tetraodontiforms, an order of fishes which have scales modified into armoured plates. While these species may not be fast swimmers, their bodies are highly modified to counter predation, which may allow it them forage in exposed areas.

- 65 -

a) 0.7 c) 0.7 0.6 0.6

0.5 0.5 ρ TL1 ρ TL1 0.4 0.4 TL2 TL2 0.3 0.3 TL3 TL3 Spearman Spearman Spearman 0.2 0.2 TL4 TL4 0.1 0.1

0.0 0.0 1B:1H 3B:3H 6B:6H 12B:12H 1U:3H 2U:6H 4U:12H

b) 0.7 d) 0.7 -

0.6 0.6 6 6

0.5 0.5 ρ TL1 ρ TL1 0.4 0.4 TL2 TL2 0.3 0.3 TL3 TL3 Spearman Spearman Spearman 0.2 0.2 TL4 TL4 0.1 0.1

0.0 0.0 1B:1H 3B:3H 6B:6H 12B:12H 1U:3H 2U:6H 4U:12H

Figure 5: Mean correlation (|ρ| ± SE) between abundance and PCO axes based on habitat variables for different levels of habitat aggregation and species in four body size classes (TL1, TL2, TL3 and TL4 – see ‘Methods’ for details). Correlations are for (a,c) PCO axis 1 and (b,d) PCO axis 2 , and data collected by (a,b) BRUV and (c,d) UVC. Dashed lines are present only to facilitate interpretation and are not meant to imply a continuous response.

- 66 -

In this research, we used maximum reported length rather than in situ estimations of fish length, which could mask ontogenetic changes in fish-habitat associations. For example, the only TL4 species strongly associated with 300 m of habitat data using BRUV was Chrysophrys auratus (Table 3). On BRUV tapes, the majority of C. auratus observed were juveniles (Colton unpubl. data), meaning that these individuals are not at the maximum length for the species and so may experience habitat at a finer scale than their length category would suggest. This could explain why the interaction between length and grain was only significant for the UVC data on PCO1 (Table 4), and suggests that using in situ measurements could strengthen the trends seen in Fig. 5. Unfortunately, diver estimates of fish size are frequently biased and unreliable (e.g., Edgar et al. 2004), which is why we opted to use maximum reported lengths in this research. Stereo-video units have shown promise in accurately measuring fish size (Cappo et al. 2003), and could provide a future method by which to investigate size-dependent fish-habitat associations.

In this study, fish and habitat surveys were conducted separately, leaving us to draw general conclusions about fish-habitat associations that were not based on concurrent observations. While most of the species observed in this study move little as adults, they undoubtedly undertake diel movements either to feed or as a result of inter- or intra-specific interactions. In addition, some of these species may exhibit diver-averse behaviour or could be especially attracted to the BRUV bait (Colton and Swearer 2010), with the result that species could have been observed in areas where they do not usually occur. This may explain why more species were observed in shallow simple habitats using BRUV than UVC (Fig. 3): they could have been attracted away from their usual habitat or refuge by the bait.

Understanding the scale at which species respond to their environment is vital to conservation and management. Species are increasingly threatened by human encroachment, which fragments natural areas. One approach to mitigation is to establish protected areas, a technique that has been used in terrestrial environments for centuries but has only relatively recently been considered an essential conservation tool in marine ecosystems (Ward et al. 2001). Two of the questions facing managers who implement protected areas, be they terrestrial or marine, is how to site them to facilitate connectivity (through movement of larvae and/or adults) and - 67 -

how large they need to be in order to sustain populations. Addressing these questions necessitates an understanding of species’ functional grain – or, in other words, how and why species vary in the spatial scales at which they respond to habitat heterogeneity. The results of this study highlight the importance of considering the scale at which species relate to their environment (Wiens 1989, Sale 1998) when making management decisions. This research suggests that fishes may respond to their environment at larger scales than previously considered. Development of a more mechanistic understanding of fish-habitat associations, beyond just body size, will greatly improve our ability to protect and manage marine resources and to predict ecosystem responses to environmental change.

LITERATURE CITED Anderson MJ, Millar RB (2004) Spatial variation and effects of habitat on temperate reef fish assemblages in northeastern New Zealand. Journal of Experimental Marine Biology and Ecology. 305:191-221. Anderson TW (1989) The relationship between habitat structure, body size and distribution of fishes at a temperate artificial reef. Bulletin of Marine Science 44:681-697. Baguette M, Van Dyck H. (2007) Landscape connectivity and animal behavior: functional grain as a key determinant for dispersal. Landscape Ecology 22:1117-1129. Beukers JS, Jones GP (1998) Habitat complexity modifies the impact of piscivores on a coral reef fish population. Oecologia 114:50-59. Biedermann R (2003). Body size and area-incidence relationships: is there a general pattern? - Global Ecology and. Biogeography 12:381-387. Boscolo D, Metzger JP (2009) Is bird incidence in Atlantic forest fragments influenced by landscape patterns at multiple scales? Landscape. Ecology. 24:907-918. Boyce MS (2006) Scale for resource selection functions. Diversity and Distribution 12:269- 276. Brown JH (1984) On the relationship between abundance and distribution of species. American Naturalist 124:255-279. Cappo M,.De'Ath G, Speare P (2007) Inter-reef vertebrate communities of the Great Barrier Reef Marine Park determined by baited remote underwater video stations. Marine Ecology-Progress Series 350:209-221. Cappo M, Harvey E, Malcolm H, Speare P (2003) Potential of video techniques to monitor diversty, abundance and size of fish in studies of marine protected areas. In: Beuner JP, Grant A, Smith DC, Mahon D. (eds.) Aquatic Protected Areas: What works best and how do we know? Proceedings of the World Congress on Aquatic Protected Areas, Cairns, Australia; August 2002, pp. 455-464. Chatfield BS, Van Niel KP, Kendrick GA, Harvey ES (2010) Combining environmental gradients to explain and predict the structure of demersal fish distributions. Journal of Biogeography 37:593-605. Choat JH, Ayling AM (1987) The relationship between habitat structure and fish faunas on New Zealand reefs. - Journal of Experimental Marine Biology and Ecology 110:257-284. Clarke KR, Warwick RM (2001) Change in Marine Communities: An Approach to Statistical Analysis and Interpretation. Primer-E Ltd., Plymouth U.K.

- 68 -

Colton MA, Swearer SE (2010) A comparison of two survey methods: differences between underwater visual census and baited remote underwater video. Marine Ecology- Progress Series 400:19-36. Connell SD, Jones GP (1991) The influence of habitat complexity on postrecruitment processes in a temperate reef fish population. Journal of Experimental Marine Biology and Ecology 151:271-294. Dahlgren CP, Eggleston DB (2000) Ecological processes underlying ontogenetic habitat shifts in a coral reef fish. Ecology 81:2227-2240. Denny CM (2005) Distribution and abundance of labrids in northeastern New Zealand: the relationship between depth, exposure and pectoral fin aspect ratio. Environmental Biology of Fishes 72:33-43. Eagle JV, Jones GP, McCormick ME (2001) A multi-scale study of the relationships between habitat use and the distribution and abundance patterns of three coral reef angelfishes (Pomacanthidae). Marine Ecology- Progress Series 214:253-265. Edgar GJ, Barrett NS, Morton AJ (2004) Biases associated with the use of underwater visual census techniques to quantify the density and size-structure of fish populations. Journal of Experimental Marine Biology and Ecology 308:269-290. Eschmeyer WN, Fricke R (2009) Catalog of Fishes electronic version (9 September 2009). http://research.calacademy.org/ichthyology/catalog/fishcatmain.asp. Feist BE, Steel EA, Jensen DW, Sather DND (2010) Does the scale of our observational window affect our conclusions about correlations between endangered salmon populations and their habitat? Landscape Ecology 25:727-743. Foster SA (1999) The geography of behaviour: an evolutionary perspective. Trends in Ecology and Evolution 14:190-195. Fulton CJ, Bellwood DR (2004) Wave exposure, swimming performance, and the structure of tropical and temperate reef fish assemblages. Marine Biology 144: 429-437. Fulton CJ, Bellwood DR, Wainwright PC (2001) The relationship between swimming ability and habitat use in wrasses (Labridae). Marine Biology 139:25-33. Gomon MF, Bray D, Kuiter R (2008) Fishes of Australia's Southern Coast. Reed New Holland, Sydney. Grinnell J (1917) Field tests of theories concerning distributional control. American Naturalist 51:115-128. Grober-Dunsmore R, Frazer TK, Beets JP, Lindberg WJ, Zwick P, Funicelli NA (2008) Influence of landscape structure, on reef fish assemblages. Landscape Ecology 23:37-53. Guidetti P, Bussotti, S (2000) Nearshore fish assemblages associated with shallow rocky habitats along the southern Croatian coast (eastern Adriatic sea). Vie et Milieu - Life and Environment 50:171-176. Holland JD, Fahrig L, Capuccino N (2005) Body size affects the spatial scale of habitat-beetle interactions. Oikos 110:101-108. Holling CS (1992) Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62:447-502. Huettmann F, Diamond AW (2006) Large-scale effects on the spatial distribution of seabirds in the Northwest Atlantic. Landscape Ecology 21:1089-1108. Jenkins SH (1981) Common patterns in home range body size relationships of birds and mammals. American Naturalist 118(1):126-128. Johansen JL, Bellwood DR, Fulton CJ (2008) Coral reef fishes exploit flow refuges in high- flow habitats. Marine Ecology- Progress Series 360:219-226. Kohler KE, Gill SM (2006) Coral Point Count with Excel extensions (CPCe): A Visual Basic program for the determination of coral and substrate coverage using random point count methodology. - Computers and Geosciences 32:1259-1269. Kuiter RH (2000). Coastal Fishes of South-eastern Australia. - Gary Allen Pty Ltd, Sydney. Magnhagen C, Borcherding J (2008) Risk-taking behaviour in foraging perch: does predation pressure influence age-specific boldness? Animal Behaviour 75:509-517. McClanahan TR, Shafir SH (1990) Causes and consequences of sea urchin abundance and diversity in Kenyan coral reef lagoons. Oecologia 83:362-370.

- 69 -

Meyer CB, Thuiller W (2006) Accuracy of resource selection functions across spatial scales. Diversity and Distributions 12:288-297. Murakami M, Hirao T, Iwamoto J, Oguma H (2008) Effects of windthrow disturbance on a forest bird community depend on spatial scale. Basic and Applied Ecology 9:762-770. Pérez-Matus A, Ferry-Graham LA, Cea A, Vásquez JA (2007) Community structure of temperate reef fishes in kelp-dominated subtidal habitats of northern Chile. Marine and Freshwater Research 58:1069-1085. Reaka ML (1980) Geographic range, life history patterns, and body size in a guild of coral- dwelling Mantis shrimps. Evolution 34:1019-1030. Ritchie ME, Olff H (1999) Spatial scaling laws yield a synthetic theory of biodiversity. Nature 400:557-560. Roland J, Taylor PD (1997) Insect parasitoid species respond to forest structure at different spatial scales. Nature 386:710-713. Sagnes P, Champagne JY, Morel R (2000) Shifts in drag and swimming potential during grayling ontogenesis: relations with habitat use. Journal of Fish Biology 57:52-68. Sale PF (1998) Appropriate spatial scales for studies of reef-fish ecology. Australian Journal of Ecology 23:202-208. Shoji J, Sakiyama K, Hori M, Yoshida G, Hamaguchi M (2007) Seagrass habitat reduces vulnerability of red sea bream Pagrus major juveniles to piscivorous fish predator. Fisheries Science 73:1281-1285. Slusarczyk M, Pinel-Alloul B (2010) Depth selection and life history strategies as mutually exclusive responses to risk of fish predation in Daphnia. Hydrobiologia 643:33-41. Swihart RK, Slade NA, Bergstrom BJ (1988) Relating body size to the rate of home range use in mammals. Ecology 69(2):393-399. Underwood AJ, Chapman MG, Crowe TP (2004) Identifying and understanding ecological preferences for habitat or prey. Journal of Experimental Marine Biology and Ecology 300:161-187. Ward TJ, Heinemann D, Evans N (2001) The Role of Marine Reserves as Fisheries Management Tools: A Review of Concepts, Evidence and International Experience. Department of Agriculture, Fisheries & Forestry, Canberra Australia. Wellnitz TA, Poff NL, Cosyleon G, Steury B (2001) Current velocity and spatial scale as determinants of the distribution and abundance of two rheophilic herbivorous insects. Landscape Ecology 16:111-120. Werner EE (1992) Individual behavior and higher-order species interactions. American Naturalist 140: S5-S32. Wiens JA (1989) Spatial scaling in ecology. Functional Ecology 3: 385-397. Willis TJ (2001) Visual census methods underestimate density and diversity of cryptic reef fishes. Journal of Fish Biology 59:1408-1411. Willis TJ, Anderson MJ (2003) Structure of cryptic reef fish assemblages: relationships with habitat characteristics and predator density. Marine Ecology-Progress Series 257:209-221.

- 70 -

CHAPTER THREE

Locating faunal breaks in the nearshore fish assemblage of Victoria, Australia

ABSTRACT Despite the potential for dispersal that many marine organisms possess, marine communities are frequently biogeographically structured. The marine environment of south-eastern Australia contains several features that have been implicated as causal factors in species’ range terminations, and previous research in Victoria has suggested it is the site of disjunctions for several marine taxa. However, little research has focused on nearshore rocky reef ichthyofauna, and the location of the disjunction remains debated. Using fish density measured by two methods, underwater visual census (UVC) and baited remote underwater video (BRUV), at six locations along the open coast of Victoria, we examined (1) whether there is sufficient concordance among species to indicate the presence of a faunal break for Victoria’s ichthyofauna; and (2) where any such breaks occur. Differences in assemblage composition between locations were tested with analyses of similarity (ANOSIMs) and examination of residuals from regressions of pair-wise dissimilarities against coast-line distance. Data collected using UVC indicated the presence of a large faunal break in the vicinity of Ninety Mile Beach, and a second break between Capes Conran and Howe, suggesting that contemporary habitat discontinuities, current flow and/or water temperature may be important factors structuring communities in this region.

KEY WORDS: South-eastern Australia; East Australian Current; Southern Australian Current; dispersal; colonization; climate change

INTRODUCTION Effective management and conservation require an understanding of species’ distributions in general and biogeographic structure in particular (Lourie & Vincent 2004). Increasingly, managers and conservationists also need to understand how species’ ranges may shift in response to climate change. Identification of faunal breaks can provide insight into the factors that determine species’ ranges, which are ultimately limited by a species’ ability to disperse to and colonize an area (Myers

- 71 -

1997). Knowing the limitations to species’ ranges can facilitate predicting how species may respond to changing environmental conditions.

Discrete faunal boundaries have been identified in numerous coastal marine communities throughout the world (Horn & Allen 1978, Murray & Littler 1981, Pondella et al. 2005). Several physical factors have been found to cause disjunctions in the marine environment, including habitat discontinuities (Riginos & Nachman 2001, Pelc et al. 2009), circulation patterns (Gaylord & Gaines 2000, Pelc et al. 2009), and water temperature (Pondella et al. 2005). Species-specific attributes interact with these physical factors to delineate ranges. For reef-associated organisms, which are relatively sedentary as adults, dispersal and colonization primarily occur during the pelagic larval and benthic juvenile stages, respectively. Although physical barriers can limit larval dispersal, hydrodynamic barriers caused by flow can occur in areas of continuous habitat (Gaylord & Gaines 2000). Flow can also facilitate dispersal (Booth et al. 2007), even though larvae have been shown to have impressive swimming abilities and sensory capacities (Leis & McCormick 2002). Whether larvae survive to successfully colonize will depend on their ability to locate benthic habitat and the suitability of that habitat, which will depend upon the match between a species’ phenotype and the environment (Figueira & Booth 2010, Marshall et al. 2010, Shima & Swearer 2010), and by multi-species interactions (Wethey 2002, Sexton et al. 2009).

The state of Victoria in south-eastern Australia is a complex and dynamic region (Fig. 1) in which several factors occur that can influence both dispersal and colonization in temperate reef communities. First, there exists a contemporary discontinuity in subtidal rocky reef habitat that stretches over 300 km from Ninety Mile Beach to the mangroves of eastern Wilsons Promontory (Fig. 1). Second, during historical glacial periods, Victoria was connected to Tasmania via the Bassian Isthmus in what is currently Bass Strait (Lambeck & Chappell 2001), which created an absolute barrier to east-west dispersal for species unable to tolerate the cold waters at the southern end of the Isthmus. Third, Victoria is a convergence zone for at least three distinct currents (Fig. 1). The Southern Australian Current (SAC) originates in Western Australia as the warm-water Leeuwin Current, gathering high salinity waters from the

- 72 -

Figure 1: A map showing the approximate location of major circulation patterns and three zoogeographical provinces in southern Australia (after Whitely 1932, O’Hara & Poore 2000, Waters & Roy 2003, Ridgway & Condie 2004). Inset shows the location of two major landmarks, Wilsons Promontory and Ninety Mile Beach, and approximate circulation currents in Bass Strait (after Gibbs 1991) using the following abbreviations: SAC = South Australian Current; ZC = Zeehan Current, SAW = Subantarctic water; BSW = Bass Strait water, and EAC = East Australian Current. The dashed line between the SAC and the ZC indicates likely connectivity between these currents (Ridgway and Condie 2004). Acronyms in grey are locations of study sites, with AB = Apollo Bay, BH = Barwon Heads, PW = west coast of Wilsons Promontory, PE = east coast of Wilsons Promontory, CC = Cape Conran and CH = Cape Howe.

- 73 -

Figure 2: Number of surveys of each type conducted at each of six locations.

Great Australian Bight as it flows eastward, to eventually turn southward in western Bass Strait to become the Zeehan current, which flows as far as the south-eastern coast of Tasmania (Ridgway & Condie 2004). The eastern part of Victoria is influenced by the East Australian Current (EAC), which brings warm waters south from the South Equatorial Current (Ridgway & Dunn 2003). In the centre of the state lies Bass Strait, a shallow and fairly stagnant body of water (Sandery & Kampf 2005) with a tendency to eastward flow (Fandry 1983, Gibbs 1991). Bass Strait receives input both from nutrient-rich, cold subantarctic waters and the warm, high-salinity Southern Australian Current (Gibbs 1991, Ridgway & Condie 2004). These complex circulation patterns and their attendant variable temperature gradients, which are occasionally extreme (Gibbs 1991), have the potential to limit both colonization and dispersal (Gaylord & Gaines 2000).

The first description of Australian zoogeography was developed by Whitely (1932) who described three provinces that overlap in south-eastern Australia: the Flindersian, Maugean and Peronian (Fig. 1). A more complete description of bioregionalisation, based on fish composition and richness, was developed by Lyne et al. (1996), which included a detailed description of the zoogeographical provinces and areas of overlap in Victoria. As the authors note, describing Victoria’s biogeography is made challenging by the presence of both east–west and north–south influences, necessitating that this area be analysed in two dimensions. As the authors describe, the fish fauna in Victoria is comprised of elements from: (1) the South Eastern Zootone, which extends from Sydney to Wilsons Promontory; (2) the Bass Strait Province, which extends from near Apollo Bay to Wilsons Promontory and south to Tasmania; (3) the Tasmanian Province, which encircles the island of Tasmania, with species - 74 - from this province extending westward and northward; and (4) the Western Bass Strait Zootone, which extends from the South Australian Gulfs to near Apollo Bay. Considering this complexity, it is unsurprising that many species’ ranges terminate in Victoria, including those of warm temperate species from the east coast, cool temperate species from the south east region, cold water species from Tasmania, and species from central southern Australia.

While several studies have identified Victoria as containing a faunal break, the reported location of the disjunction varies. Wilsons Promontory has long been viewed as a transition zone (Whitely 1932, Hough & Mahon 1994), and recent studies have demonstrated that it is a site of disjunction for several taxa, including marine echinoderms and decapods (O'Hara & Poore 2000), littoral gastropods (Waters et al. 2005), a scyphozoan (Dawson 2005), and several fish species (Kuiter 2000; Gomon et al. 2008). Other studies place the location of the disjunction at Ninety-Mile Beach, e.g., for Paraplesiops spp. (Hutchins 1987), rocky intertidal invertebrates (Hidas et al. 2007), and some fish species (Kuiter 2000, Gomon et al. 2008). Still other studies have found taxa to have phylogeographic structure that is best explained by the emergence of the Bassian Isthmus (Waters & Roy 2003, Waters et al. 2004), or with range expansions and contractions associated with cooling temperatures during glacial periods (Burridge 2000). Finally, at least one study has found that contemporary population genetic structure is best explained as the reinforcement of historical factors by present-day flow (Waters 2008). Despite the complexity of this region, little research has focused on the distribution of its nearshore fish assemblages.

In this study, we explored the biogeography of the nearshore rocky reef ichthyofauna of Victoria. While there is evidence to suggest that Victoria is the site of range terminations for several fish species (Kuiter 2000, Gomon et al. 2008), it is unclear whether these terminations co-occur for multiple fishes at a single geographical location. In addition to exploring evidence for inter-specific range boundaries, we also explored whether Wilsons Promontory or Ninety-Mile Beach is associated with larger changes in community structure.

- 75 -

MATERIALS AND METHODS Data collection Surveys were conducted on nearshore rocky reefs in four locations along the coast of Victoria in south-eastern Australia between December 2007 and June 2008, and at two additional locations between March and May 2009 (Fig. 2). At each location, data were collected using both underwater visual census (UVC) and baited remote underwater video (BRUV) (methods detailed in Colton & Swearer 2010) (Fig. 2). An exploration of habitat variation across the region and its effects on fish assemblage composition is discussed further in Chapter Three. These two methods were chosen in order to capture as much of the fish assemblage as possible, as our previous research revealed that UVC and BRUV together provided a more complete estimate of community composition than either method alone (Colton & Swearer 2010). Using UVC, we measured fish density (m-2), while using BRUV provided us with a measure of relative density called MaxN. MaxN is a species-specific measure of the maximum number of individuals observed in a one-second interval on a BRUV tape.

Data analysis Data were removed of all individuals that were not identified to the species level, with the exception of a putative species Trygonorrhina sp. A (Gomon et al. 2008). To reduce noise in the data associated with poorly sampled species, only species observed more than once per method were included in the analyses. In addition, it was only sometimes possible to identify to species individuals in the genus Upeneichthys as morphometric differences between U. lineatus and U. vlamingii species are slight (Gomon et al. 2008). Without the ability to consistently identify individuals to the species level, the utility of Upeneichthys spp. to distinguish between locations diminished, especially because Upeneichthys spp. is ubiquitous. Therefore, all records of individuals in the genus Upeneichthys were removed from the analysis. Taxonomic nomenclature follows Eschmeyer & Fricke (2009) and species distribution information was based on field guides (Table 1).

Data collected using the two methods were combined to determine species turnover between neighbouring locations and the maximum range of each species. Turnover was calculated as the total number of species with ranges that ended in either of two neighbouring locations (sensu O'Hara & Poore 2000). Species that occurred at all

- 76 - sample locations on either side of a pair of neighbouring locations were categorized as having a range termination, and species which were observed at all locations were categorized as “ubiquitous”.

Statistical analyses were performed on the abundance data (density measured by UVC and MaxN measured by BRUV) collected by each method separately because this and other research (Colton & Swearer 2010) revealed significant differences between methods. In these analyses, we chose to examine density and MaxN rather than simple presence–absence data because doing so can elucidate patterns that would be overlooked by consideration of species’ ranges alone (Waters 2008). Data were fourth-root transformed and analyses conducted on a Bray–Curtis resemblance matrix. Multivariate differences in species assemblages between locations were visualized using non-metric multi-dimensional scaling (MDS) plots (Primer–E v.6, Clarke & Warwick 2001). Testing of differences between locations was accomplished using the analysis of similarity (ANOSIM) routine in Primer–E with 9999 permutations. ANOSIM is analogous to univariate ANOVA but operates on a resemblance matrix. The test statistic generated by an ANOSIM is an R-value that measures the amount of dissimilarity between a priori groups of samples. R ranges from 0, no difference between groups, to +1, perfect dissimilarity between groups. Finally, a formal test of whether locations clustered together was performed using the similarity profile (SIMPROF) routine in Primer–E. SIMPROF is an unconstrained ordination by permutation test for determining the number of significant clusters. Clustering was accomplished by applying the group average method to mean MaxN (BRUV) and density (UVC) for all species at each location, and significance assessed using 9999 permutations.

We examined dissimilarity between locations as a function of coastline distance, measured using Google Earth, in order to explore specific hypotheses. If there was little biogeographic structure to the region we would expect all pair-wise dissimilarities between locations to be equal. In contrast, if the region contained a biogeographic disjunction, we would expect a dramatic change in assemblage composition over a small distance. To explore these hypotheses, we examined ANOSIM dissimilarities between pairs of locations as a function of coastline distance

- 77 -

Table 1: Species observed at each location by b = BRUV and u = UVC. Superscripts indicate the following: the five most abundant species for each location with * = BRUV and § = UVC; ubiquitous species with † = BRUV and ‡ = UVC. Taxonomy follows Eschmeyer & Fricke (2009). Distribution information is based on information in field guides (Kuiter 2000, Edgar 2005, Gomon et al. 2008) with rare occurrences at range edges not included. Widespread southern species are present in most of southern Australia, e.g., from Queensland to Western Australia; Southern species are primarily Tasmanian that occur on the mainland in the vicinity of Bass Strait; South-central species occur only on the south coast including Tasmania, Victoria, South Australia and/or Western Australia; Eastern are species that occur on the east coast including eastern Victoria but not Wilsons Promontory; Eastern in Victoria are species which are widespread across southern Australia and yet in Victoria are only observed in the east; and South-eastern species occur on the east coast and at least as far as Wilsons Promontory but no further than South Australia. The lone South- western species is most abundant in southern Western Australia and also occurs in the vicinity of Port Philip Bay. Locations are abbreviated as in Fig.1. Distribution Species Family AB BH PW PE CC CH Widespread southern Acanthaluteres vittiger† Monacanthidae bu bu bu bu b bu Widespread southern Aracana aurita Aracanidae bu bu u Widespread southern Aulopus purpurissatus Aulopidae bu b u -

‡ * 78 Widespread southern Cephaloscyllium laticeps Scyliorhinidae bu bu bu u bu bu

‡ -

Widespread southern Cheilodactylus nigripes Cheilodactylidae bu bu bu bu bu Widespread southern Chrysophrys auratus Sparidae bu* b b b bu Widespread southern Dactylophora nigricans Cheilodactylidae bu b u b Widespread southern Dasyatis brevicaudata Dasyatidae b b b b Widespread southern Dinolestes lewini†‡ Dinolestidae bu bu bu* bu bu bu Widespread southern Diodon nicthemerus‡ Diodontidae u u u u u u Widespread southern Dotolabrus aurantiacus Labridae bu u bu bu u b Widespread southern Enoplosus armatus†‡ Enoplosidae bu bu bu bu bu bu Widespread southern Eubalichthys bucephalus Monacanthidae bu Widespread southern Eubalichthys mosaicus Monacanthidae b u bu b Widespread southern Eupetrichthys angustipes Labridae u bu u bu u Widespread southern Girella zebra†‡ Kyphosidae bu bu§ bu bu bu bu Widespread southern Haletta semifasciata Odacidae bu b Widespread southern Heterodontus portusjacksoni† Heterodontidae b b b b b bu

- 78 -

Table 1 Continued Distribution Species Family AB BH PW PE CC CH Widespread southern Heteroscarus acroptilus Odacidae bu b bu bu u Widespread southern Hypoplectrodes nigroruber Serrandiae u u Widespread southern Lepidotrigla vanessa Trigliidae u Widespread southern Lotella rhacina Moridae bu u Widespread southern Meuschenia flavolineata Monacanthidae u bu§ bu bu b u Widespread southern Meuschenia freycineti† Monacanthidae bu bu bu bu b bu Widespread southern Meuschenia hippocrepis Monacanthidae bu*§ bu u b Widespread southern Meuschenia scaber Monacanthidae u bu bu Widespread southern Meuschenia venusta Monacanthidae bu u Widespread southern Mustelus antarcticus Triakidae b b b Widespread southern Myliobatis australis Myliobatidae bu b b b b b Widespread southern Nemadactylus macropterus Cheilodactylidae u b -

79 Widespread southern Neoodax balteatus Odacidae u bu u

- Widespread southern Olisthops cyanomelas†‡ Odacidae bu bu bu§ bu bu bu

Widespread southern Parequula melbournensis Gerreidae b bu bu bu bu Widespread southern Parma victoriae Pomacentridae bu bu bu bu Widespread southern Pempheris multiradiata‡ Pempherididae u§ bu§ u u u u Widespread southern Pentaceropsis recurvirostris‡ Pentacerotidae u u bu u bu bu Widespread southern Pictilabrus laticlavius†‡ Labridae bu bu bu bu bu bu Widespread southern Pseudolabrus psittaculus† Labridae bu b bu bu bu* bu Widespread southern Pseudophycis barbata Moridae b b b Widespread southern Rexea solandri Gempylidae b Widespread southern Scobinichthys granulatus Monacanthidae u b u Widespread southern Scorpis aequipinnis Scorpididae bu bu* bu§ bu u Widespread southern Seriola lalandi Carangidae b b Widespread southern Sillaginodes punctatus Sillaginidae b* b bu b* Widespread southern Siphonognathus beddomei Odacidae u Widespread southern Sphyraena novaehollandiae Sphyraenidae b b b b bu

- 79 -

Table 1 Continued Distribution Species Family AB BH PW PE CC CH Widespread southern Tilodon sexfasciatum Kyphosidae u bu bu bu Widespread southern Trachurus declivis Carangidae u u Widespread southern Trachurus novaezelandiae Carangidae b b* u b* bu*§ Widespread southern Trygonorrhina fasciata Rhinobatidae bu b Southern Eubalichthys gunnii Monacanthidae bu Southern Meuschenia australis Monacanthidae u bu South-central Aplodactylus arctidens†‡ Aplodactylidae bu bu bu bu bu bu South-central Aracana ornata Aracanidae u bu South-central Caesioperca rasor†‡ Serrandiae bu§ bu* bu*§ bu*§ bu bu South-central Eeyorius hutchinsi Moridae b b South-central Platycephalus bassensis Platycephalidae b b b b South-central Pseudophycis bachus Moridae u

- *§

80 South-central Trachinops caudimaculatus Plesiopidae bu u

- Eastern Acanthistius ocellatus Serrandiae b bu

Eastern Allomycterus pilatus Diodontidae b Eastern Aplodactylus lophodon Aplodactylidae bu bu Eastern Cheilodactylus fuscus Cheilodactylidae bu Eastern Chromis hypsilepis Pomacentridae bu§ Eastern Coris sandageri Labridae u Eastern Dicotylichthys punctulatus Diodontidae u u u Eastern Notolabrus gymnogenis Labridae u bu bu Eastern Pempheris compressa Pempherididae u u Eastern Suezichthys aylingi Labridae u Eastern Trachinops taeniatus Plesiopidae u§ Eastern in Victoria Gymnothorax prasinus Muraenidae b bu Eastern in Victoria Ophthalmolepis lineolata Labridae bu bu* bu* South-eastern Achoerodus viridis Labridae b bu bu South-eastern Atypichthys strigatus Kyphosidae u bu*§ bu§ bu§ bu*§

- 80 -

Table 1 Continued Distribution Species Family AB BH PW PE CC CH South-eastern Caesioperca lepidoptera Serrandiae bu bu§ bu*§ bu*§ South-eastern Cheilodactylus spectabilis‡ Cheilodactylidae bu u bu bu bu bu South-eastern Chironemus marmoratus Chironemidae bu South-eastern Girella elevata Kyphosidae u bu South-eastern Hypoplectrodes maccullochi Serrandiae bu bu South-eastern Latridopsis forsteri Latridae bu bu bu b bu South-eastern Nemadactylus douglasii Cheilodactylidae b bu bu South-eastern Notolabrus fucicola†‡ Labridae bu*§ bu bu* bu bu bu South-eastern Notolabrus tetricus†‡ Labridae bu*§ bu*§ bu*§ bu*§ bu*§ bu South-eastern Optivus agastos Trachichthyidae u u South-eastern Parma microlepis Pomacentridae u bu bu§ bu South-eastern Scorpis lineolata Scorpididae bu bu bu*§ bu§ bu* -

81 South-eastern Tetractenos glaber Tetraodontidae bu b bu b

- South-eastern Trygonorrhina sp. A Rhinobatidae b b

South-eastern Urolophus cruciatus Urolophidae bu u South-western Meuschenia galii Monacanthidae u bu

- 81 -

using linear regression conducted in SPSS (v.16.0). In order to identify locations between which large differences occurred that were not a function of geographic distance, we plotted regression residuals, computed as the observed value minus the value predicted by the regression equations.

Finally, for species which showed significant differences in abundance between neighbouring locations (for either BRUV or UVC), we tested whether species with similar distributions exhibited consistent changes (i.e., increases or decreases) in abundance between locations using χ2 tests.

RESULTS Species turnover and range disjunctions A total of 91 species were included in these analyses, of which 78 were observed using BRUV and 80 using UVC (Table 1). Twenty-one species were ubiquitous, of which nine were observed by both methods, 14 only by BRUV, and 16 only by UVC. Some eastern species, such as Dicotylichthys punctulatus and Pempheris compressa, were observed as far west as Wilsons Promontory, while others, such as Trachinops taeniatus and Suezichthys aylingi, were only recorded at Cape Howe. Most south- eastern species ranged as far west as Wilsons Promontory, though Scorpis lineolata was observed at Barwon Heads and Hypoplectrodes maccullochi was only recorded west to Cape Conran. The primarily Tasmanian species Meuschenia australis and Eubalichthys gunnii were only recorded in Bass Strait at Barwon Heads and Wilsons Promontory.

Species turnover was highest between the east coast of Wilsons Promontory and Cape Conran, and lowest between Apollo Bay and Barwon Heads (Table 2). Thirty-two species (35%) exhibited evidence of a range disjunction and 23% were ubiquitous. The majority of range terminations occurred between the east coast of Wilsons Promontory and Cape Conran (n = 12), with a similar number occurring between Cape Conran and Cape Howe (n = 10). A total of 10 species (11%) had range breaks between the other three locations.

The region did not appear to be equally permeable to dispersal, and there was some indication that species’ distributions were more limited in westward than eastward - 82 -

Table 2: Species turnover and the number of species exhibiting range disjunctions between neighbouring locations. Species’ distribution described by a left arrow indicates that the species occurs at that and all locations to the west, and a right arrow indicates number of species occurring at that and all locations to the east. Locations are abbreviated as in Fig. 1. Locations Turnover Distribution No. of species AB / BH 19 BH-> 2 <-AB 1 BH / PW 21 PW -> 2 <-BH 1 PW / PE 25 PE -> 4 <- PW 0 PE / CC 32 CC -> 8 <-CC 3 Ubiquitous 21 dispersal. For example, 100% of the species with a range ending at Wilsons Promontory occurred to the east of the Promontory, 80% of the species with a break between Cape Conran and the east coast of Wilsons Promontory occurred at Cape Conran, and 70% of the species with a break between Cape Conran and Cape Howe occurred to the east of Cape Howe (Table 2). Similarly, the breaks between Barwon Heads and both the west coast of Wilsons Promontory and Apollo Bay also appeared to act more as a barrier to westward than to eastward dispersal.

Assemblage composition by location MDS plots revealed differences between locations for both BRUV and UVC data (Fig. 3). In both data sets, the eastern locations, Cape Conran and Cape Howe, were clearly differentiated from the rest of the state. In the BRUV data, the samples from the other four locations overlapped. In contrast, the UVC data suggest that samples from both the east and west coasts of Wilsons Promontory were intermediate between those from the east, Cape Conran and Cape Howe, and those from the west, Apollo Bay and Barwon Heads. In addition, there is some indication in the UVC data that the east coast of Wilsons Promontory was more similar to Cape Conran and Cape Howe, and the west coast of Wilsons Promontory more similar to Apollo Bay and Barwon Heads.

Several pairs of congeneric species appeared to be separated into eastern and western components (further phylogenetic work is required to ascertain whether these are sister species). Caesioperca rasor was observed at all locations but was more abundant in the west, while C. lepidoptera was present only in the east. Parma

- 83 -

(a)

(b)

Figure 3: MDS plots showing differences in assemblage composition for data collected by (a) BRUV, and (b) UVC. Locations are abbreviated as in Fig. 1. microlepis was found only in eastern samples and P. victoriae in western samples, though the species overlapped at Wilsons Promontory. There was also significant overlap in the occurrence of Scorpis aequipinnis and S. lineolata, though the former was more abundant in the west and not observed at Cape Howe, and the latter more abundant in the east and not observed at Apollo Bay. Trachinops caudimaculatus was

- 84 -

only observed at Baron Heads and the west coast of Wilsons Promontory, while T. taeniatus was only recorded at Cape Howe. ANOSIM revealed an overall significant effect of location for BRUV (R = 0.44, p = 0.0001) and UVC (R = 0.66, p = 0.0001), and all pair-wise comparisons between locations were significant for both methods (Table 3). However, there was a wide range in the strength of R. Examining pair-wise comparisons for only neighbouring locations, BRUV data revealed little differences across the five comparisons (mean = 0.24 ± 0.07 SD) while UVC showed both a larger range and higher values of R (mean = 0.57 ± 0.27 SD) (Table 3). The UVC data revealed the largest dissimilarity between the east coast of Wilsons Promontory and Cape Conran (R = 0.91), followed by Cape Conran and Cape Howe (R = 0.69), and Barwon Heads and the west coast of Wilsons Promontory (R = 0.63). The BRUV data showed the largest difference between Cape Conran and Cape Howe (R = 0.32), followed by the east coast of Wilsons Promontory and Cape Conran (R = 0.29), and Apollo Bay and Barwon Heads (R = 0.25).

The SIMPROF routine was used to identify significant clusters of locations. Both BRUV and UVC data revealed that the eastern locations, Cape Conran and Cape Howe, formed a significant cluster that was distinct to the other 4 locations (Fig. 4). In addition, there was non-significant structuring between the other locations that differed by methods. The BRUV data revealed that Barwon Heads was an outlier to a cluster of Apollo Bay and both coasts of Wilsons Promontory, and the east coast of Wilsons Promontory was an outlier to Apollo Bay and the west coast of Wilsons Promontory. In contrast, the UVC data revealed two non-significant clusters, one containing Apollo Bay and Barwon Heads, and the other both coasts of Wilsons Promontory. These results echo those found in MDS plots (Fig. 3), in which the UVC data suggested that the samples from Wilsons Promontory were different to those from Apollo Bay and Barwon Heads, while the BRUV data showed overlap between Apollo Bay, Barwon Heads, and Wilsons Promontory.

Dissimilarity as a function of distance In the BRUV data, ANOSIM R showed a linear relationship with coastline distance (R2 = 0.75, p < 0.0005) (Fig. 5a). Some pairs of locations had larger residuals around the regression than others (Fig. 5b). For example, Barwon Heads and the west coast of Wilsons Promontory were more similar than their distance alone would suggest, as

- 85 -

Table 3: Coastline distance (km), dissimilarity coefficients (ANOSIM R), and significance levels for pair-wise comparisons between locations. Comparisons between neighbouring locations are in bold. All comparisons are statistically significant. Locations are abbreviated as in Fig. 1. BRUV UVC Locations Distance (km) R P R P AB, BH 87 0.25 <0.0005 0.22 0.03 AB, PW 288 0.26 <0.0005 0.33 <0.005 AB, PE 306 0.49 <0.0005 0.42 <0.005 AB, CC 562 0.64 <0.0005 0.79 <0.0005 AB, CH 670 0.59 <0.0005 0.89 <0.0005 BH, PW 201 0.19 <0.005 0.63 <0.0005 BH, PE 219 0.42 <0.0005 0.68 <0.005 BH, CC 475 0.70 <0.0005 0.98 <0.005 BH, CH 583 0.68 <0.0005 1.00 <0.0005 PW, PE 18 0.15 0.02 0.40 <0.005 PW, CC 274 0.30 <0.005 0.96 <0.0005 PW, CH 382 0.58 <0.0005 0.98 <0.0005 PE, CC 256 0.29 <0.005 0.91 <0.005 PE, CH 364 0.51 <0.0005 0.96 <0.0005 CC, CH 108 0.32 <0.005 0.69 <0.0005 were Apollo Bay and the west coast of Wilsons Promontory, and Apollo Bay and Cape Howe. Some pairs of locations, such as Apollo Bay and Barwon Heads, and Barwon Heads and Cape Conran, were less similar than distance alone would suggest.

The UVC data revealed an interesting pattern in the relationship between ANOSIM R and coastline distance (R2 = 0.41, p = 0.01) (Fig. 5a). Some pair-wise comparisons were more similar than predicted by their geographic distance and some were less similar. Examination of the regression residuals clearly illustrated this (Fig. 5b). In particular, large dissimilarities between Cape Conran and Cape Howe, and the east coast of Wilsons Promontory and Cape Conran, were evident, while there was little difference between the two coasts of Wilsons Promontory.

Effect of species distribution on abundance changes between neighbouring locations Consistent changes in abundance among species with similar distributions occurred only between the east coast of Wilsons Promontory and Cape Conran (χ2 = 10.17, p = 0.017; all other comparisons: 0.087 < p > 0.786). Of species with south-central distributions, 100% were significantly more abundant at the east coast of Wilsons

- 86 -

Promontory, and 100% of species with eastern distributions and 86% of species with south-eastern distributions were significantly more abundant at Cape Conran. Changes in abundance of widespread southern species were more evenly distributed, with 62% of species significantly more abundant at the east coast of Wilsons Promontory.

BRUV vs. UVC These analyses (and previous work, see Colton & Swearer 2010) indicated that there were clear differences between the methods we used to estimate abundance. In general, BRUV appeared less sensitive to detecting biogeographic patterns than UVC. For example, an MDS plot of BRUV data showed less separation between locations than the UVC data (Fig. 3), and SIMPROF clustering revealed similar trends (Fig. 4). The UVC data alone suggested that the Wilsons Promontory locations were intermediate between locations to the east and west. There was also clear variation between the methods in their abilities to measure differences between locations. With the exception of pair-wise comparisons between Apollo Bay and the other locations, UVC found more differences between locations than BRUV as shown by higher UVC ANOSIM R-values (Fig. 5, Table 3).

DISCUSSION There were clear differences between locations as revealed by MDS plots (Fig. 3), SIMPROF clustering (Fig. 4) and ANOSIM tests (Table 3). The BRUV data exhibited a linear relationship between dissimilarity and coastline distance (Fig. 5), with little variation in dissimilarity values between pairs of neighbouring locations (Table 3). In contrast, the UVC data displayed evidence of at least one biogeographic disjunction (Fig. 5): the fish communities of Cape Conran and the east coast of Wilsons Promontory were very dissimilar (R = 0.91), suggesting a large change in assemblage composition between these locations. This result was mirrored in the χ2 results, which only found significant and consistently different changes in abundance as a function of species distribution between these two locations. Though not quite as large, the dissimilarity between Cape Conran and Cape Howe (R = 0.69) was also notable. Neither regression residuals (Fig. 5b) nor MDS plots (Fig. 3) indicated the presence of a faunal break at Wilsons Promontory.

- 87 -

(a)

(b)

Figure 4: SIMPROF clustering by location for (a) BRUV and (b) UVC data. Solid lines indicate significant clusters at P < 0.05 and dashed lines represent non-significant clusters. Locations are abbreviated as in Fig. 1.

The regressions of coastline distance against dissimilarity values for the UVC data were less straightforward than those for BRUV data (Fig. 5a). Notably, Apollo Bay

- 88 -

was more similar to all locations than would be suggested by distance alone (Fig. 5b, 6b). Several species were observed in the east region and at Apollo Bay, including Atypichthys strigatus, Latridopsis forsteri, Meuschenia scaber, and Sphyraena novaehollandiae (Table 1). Some of these species, such as A. strigatus and S. novaehollandiae are primarily east or south-eastern species (Kuiter 2000; Edgar 2005; Gomon et al. 2008), and their presence at Apollo Bay may indicate either that their range has expanded or that their ranges were previously incorrectly described. Alternatively, the similarity between Apollo Bay and the other locations could be driven by how different Barwon Heads is to the other locations, e.g., R = 1.00 for a pair-wise comparison between Barwon Heads and Cape Howe (Table 3). Barwon Heads is close to Port Philip Bay, which is known to support a distinct fish fauna (Kuiter 2000, Gomon et al. 2008); some species that are very abundant in the bay are rarely seen on the open coast, such as Trachinops caudimaculatus (pers. obs.).

Another possible explanation of the observed differences between locations is interannual variation. Cape Conran and Cape Howe were sampled in different years, as were the two coasts of Wilsons Promontory. However, as most of the species in this research are fairly sedentary as adults and have a life span greater than one year (Kuiter 2000, Gomon et al. 2008), we think that this is unlikely to offer a complete explanation. Certainly the difference in assemblage composition between Cape Conran and the east coast of Wilsons Promontory could not be explained in this manner as both these locations were sampled between March and May 2009. In addition, we found very little difference between the east and west coasts of Wilsons Promontory, which were also sampled in different years, which further supports our conclusion that interannual differences cannot fully explain the observed patterns.

Data collected by the two survey methods revealed different patterns: the BRUV data exhibited a gradation in assemblage composition in which locations that were more geographically distant were less similar than those that were geographically proximate, while the UVC data exhibited evidence of faunal disjunctions. Several studies have identified differences between UVC and BRUV (e.g., Willis & Babcock 2000, Watson et al. 2005), and in other research conducted in the same study area, we found that BRUV and UVC sampled different components of the fish fauna (Colton &

- 89 -

Swearer 2010). Notably, UVC sampled higher abundance of territorial and site- attached species, while BRUV recorded higher abundance of mobile predators. If UVC is better at surveying sedentary species and BRUV at surveying mobile species, we would expect the BRUV data to show less difference between locations than the UVC data. Indeed this is what we observed. Compared to UVC, BRUV was less capable of distinguishing between locations, which can be seen in MDS plots (Fig. 3), SIMPROF dendrograms (Fig. 4) and in dissimilarity values (Table 3).

In contrast to the BRUV data, the UVC data showed strong evidence of a biogeographic disjunction between the east coast of Wilsons Promontory and Cape Conran (Fig. 5, Table 3). The co-occurrence in this region of several factors known to influence both dispersal and colonization (i.e., converging currents, isotherms, and habitat discontinuity) makes it difficult to attribute this faunal break to a single mechanism. Indeed, while any one of these factors acting alone could influence fish distributions, it is more likely that a combination of factors structures the fish assemblages of this region, as was found for an intertidal gastropod (Waters 2008).

In this region, there is a large sandy barrier of c. 300 km in length along Ninety Mile Beach (Fig. 1), between the two locations which had the highest dissimilarity value for neighbouring locations. Other studies have found that sandy areas limit gene flow, a measure of connectivity, for rocky reef fishes. For example, Bernardi (2000) found a major phylogenetic break between populations of Embiotoca jacksoni separated by a sandy region, and Riginos & Nachman (2001) found that populations of a blennioid fish, Axoclinus nigricaudus, separated by sand were more genetically distinct that those separated by continuous rocky reef habitat. In this research, the co-occurrence of a large dissimilarity in fish assemblages with a lack of habitat in this region suggests that a sandy expanse may also structure fish communities in south-eastern Australia.

Hydrographic barriers to dispersal may also influence species’ distributions in this region. Flow has been shown to influence larval dispersal (Booth et al. 2007), despite larval behaviour (Leis & McCormick 2002). Using models, Gaylord & Gaines (2000) showed that in areas with upstream larval supply, flow can lead to discrete boundaries in species’ distributions even in the presence of continuous habitat. In this research,

- 90 -

1.0 (a)

0.9

0.8

0.7

0.6

0.5

locations 0.4

0.3 values for pairwise tests between - 0.2 BRUV 0.1 UVC

0.0 ANOSIM R 0 100 200 300 400 500 600 700 Distance (km)

0.4 (b) BRUV Less UVC similar 0.3 * 0.2 *

0.1 * * * 0.0 value residualsvalue - -0.1

-0.2

ANOSIM R -0.3

-0.4 More similar AB, PE AB, BH BH, PE PE, CC PE, CH AB, CC AB, CH BH, CC BH, CH CC, CH PW,PE AB, PW BH, PW PW,CC PW,CH

Increasing distance

Figure 5: (a) Linear regression of multivariate correlation coefficients, ANOSIM R, for pair- wise comparisons between locations as a function of coastline distance. Regression lines are dashed for BRUV and solid for UVC. Equations are y = 0.0009x + 0.135 for BRUV with R2 = 0.76 and P < 0.0005, and y = 0.001x + 0.4032 for UVC with R2 = 0.41and P = 0.01. (b) Regression residuals, with asterisks indicating pairs of neighbouring locations. Locations are abbreviated as in Fig. 1. we found evidence to suggest that the study region is differentially permeable to eastward and westward dispersal (Table 2), which may be caused by circulation patterns (Fig. 1). For example, species with southern Australian distributions were

- 91 -

more able to overcome the habitat disjunction at Ninety Mile Beach than were species with eastern distributions. However, declines in abundance of species occurring on either side of Ninety Mile beach suggest that both westward dispersal of eastern species and eastward dispersal of western species are limited by this barrier. In addition, though only four species had a range disjunction at Wilsons Promontory, 100% of these species occur to the east of the Promontory. This suggests that the predominant eastward flow in this region may play an important role in regulating species’ distributions and may, to some extent, aid eastward dispersal of organisms around the habitat-poor region of Ninety Mile Beach. A similar mechanism was proposed to explain how the larvae of Arripis trutta, a schooling species that moves considerable distances offshore, released in Western Australia may drift past the depauperate Great Australian Bight on eastward flowing currents to the west coast of Tasmania (Malcolm 1960, Ridgway & Condie 2004). While dispersal from the eastern locations to Wilsons Promontory would involve moving against predominant currents, it is certainly possible. Malcolm (1960) recaptured in South Australia a tagged Arripis trutta that had been released at the eastern end of Ninety Mile Beach, meaning this individual travelled around Wilsons Promontory and across Bass Strait. Westward movement through this region might be facilitated by the presence of a single area of rocky reef between Cape Conran and the east coast of Wilsons Promontory located at Red Bluff toward the eastern end of Ninety Mile Beach. Hidas et al. (2007) found that intertidal invertebrates that occurred on either side of Ninety Mile Beach overlapped in their distribution at Red Bluff, indicating that this site may be a stepping stone for these species. Additional surveys at this location may indicate that the same is true for fishes.

A third factor that could influence species’ distributions in this region is a paleogeographical barrier. During glacial periods, Victoria was connected to Tasmania via the Bassian Isthmus in what is currently Bass Strait (Fig. 1). This paleogeographical barrier has left its signature on the genetic structure of several taxa (Waters & Roy 2003, Waters et al. 2004) and on the contemporary distributions of a few species. In this study, several species pairs appeared to be separated into eastern and western components, e.g., Parma microlepis and P. victoriae, and Scorpis aequipinnis and S. lineolata (Table 1). These distributions could be the relic of vicariant speciation associated with the emergence of the Bassian Isthmus and cooling

- 92 -

temperatures, as has been suggested for southern Australian cirrhitoids (Burridge 2000). Other species have distributions that contain large breaks in their centre. For example, the muraenid eel Gymnothorax prasinus is found along the eastern seaboard from southern Queensland (Kuiter 2000) to eastern Victoria (this study), and along the coasts of South Australia and Western Australia (Gomon et al. 2008). The labrid Ophthalmolepis lineolata has a similar distribution (Kuiter 2000, Gomon et al. 2008), and two additional species, Trachurus declivus and Rexea solandri, are known to have different sub-populations separated in the vicinity of Bass Strait (Ward & Elliott 2001). The fact that G. prasinus has leptocephali and O. lineolata is a broadcast spawner, would suggest that these species at least have the potential to disperse to central Victoria. Perhaps it is not dispersal but colonization that limits these species’ ranges. Information about their thermal tolerances and habitat requirements could help us understand why they do not occur in central Victoria. These disjunctions also suggest the potential for incipient speciation; phylogenetic analyses would serve to further elucidate population connectivity for these and other species with discontinuous ranges.

Species that do show evidence of historical vicariance would be most likely to exhibit range disjunctions at Wilsons Promontory as this was the eastern edge of the Bassian Isthmus (Lambeck & Chappell 2001). While there was a fairly high level of species turnover associated with Wilsons Promontory, few species had ranges that terminated on either of its coasts (Table 2). Rather than being a discrete range boundary, Wilsons Promontory appears to act as a transition zone in which species from different provinces overlap. For example, the species pairs Parma microlepis and P. victoriae and Girella elevata and G. zebra co-occurred at the Promontory. P. microlepis and G. elevata are eastern species and P. victoriae and G. zebra commonly occur along the south coast (Kuiter 2000, Gomon et al. 2008). This suggests that there are occasional opportunities for dispersal between the east and west coasts of the Promontory but that they are sufficiently rare to result in strong differences in abundance on either side of the Promontory. The predominant eastward currents (Fig. 1) make the Promontory an asymmetric, leaky boundary analogous to Pt. Conception in California, USA (Wares et al. 2001). The use of abundance instead of presence– absence data provided important information about species composition. For example, though P. microlepis was observed on the west coast of Wilsons Promontory, it

- 93 -

occurred there at low density. Were our analyses to have focused only on presence/absence data, subtle gradations in assemblage composition such as these would have been overlooked.

If contemporary barriers exist then life history attributes that influence dispersal and colonization (e.g., pelagic larval duration for sedentary reef-dwelling species) are likely to be important and may explain differences among species’ ranges. However, in this research, we found little evidence to suggest a link between the potential for larval dispersal and species’ distributions. For example, G. prasinus has leptocephali, which are known to be highly dispersive, but was observed only as far west as Cape Conran. In contrast, males of the eastern pomacentrid P. microlepis guard demersal eggs and larvae are estimated to be in the plankton for only a few weeks (Curley & Gillings 2009), yet this species was observed as far as the west coast of Wilsons Promontory. Support for a general link between early life history and distribution is equivocal. For example, Booth et al. (2007) found that planktonic larval duration explained the southward extent of tropical expatriates’ ranges in south-eastern Australia, while Hidas et al. (2007) found no evidence linking dispersal potential of intertidal invertebrates with the ability to circumvent a habitat barrier in Victoria. Perhaps the shorter dispersive capacity of P. microlepis facilitates its occupation of the west coast of Wilsons Promontory by preventing larvae from being advected eastward by predominant winds and currents. Additional data on the early life history of these and other species are required to more fully explore links between life history and dispersal in south-eastern Australia.

Many of the differences in assemblage composition between Wilsons Promontory and Cape Conran, and between Cape Conran and Cape Howe could be attributable to colonization failure: species are able to disperse to the region but cannot survive once they arrive. The convergence of currents in this region results in a temperature gradient of cooler waters in the west to warmer waters in the east, with a rapid rise in temperature associated with a front in eastern Bass Strait (Gibbs 1991). The majority of species with a break in the vicinity of Ninety Mile Beach, i.e., between Wilsons Promontory and Cape Conran, are east coast species that could be limited by their ability to survive the cooler temperatures of southern Australia. Juveniles of at least one east coast species, Chromis hypsilepis, were observed at Cape Conran (pers.

- 94 -

obs.), though adults of this species were only observed at Cape Howe. The presence of juveniles and lack of adults at Cape Conran suggests that colonization failure may be the limiting factor for this species. It is quite likely that winter temperatures prevent C. hypsilepis from surviving, in a similar manner to tropical expatriates in eastern Australia whose southern range limits are largely determined by overwinter survival (Figueira & Booth 2010).

South-eastern Australia is expected to experience dramatic and potentially rapid changes in currents and sea temperatures as a result of global climate change (Figueira & Booth 2010). One study (Figueira & Booth 2010) suggests that sea surface temperatures have increased by about 1.5oC over the last 130 years, and this is already having measurable effects on the fauna of this region (Last et al. 2010). For example, the urchin Centrostephanus rodgersii has expanded its range to and around Tasmania, an expansion that is associated with strengthening of the East Australian Current (Ling et al. 2009), and numerous reef-dwelling fishes have displayed similar range expansions and increases in abundance (Last et al. 2010). In Victoria, strengthening of the EAC could result in warmer water temperatures that could allow species like C. hypsilepis to expand their range. Some range expansions may have already occurred. For example, the central southern Australian species Dactylophora nigricans and Parma victoriae were both observed as far east as the east coast of Wilsons Promontory, and Meuschenia galii, a primarily south-western Australian species that has been reported to occur in the vicinity of Port Philip Bay, was observed at Barwon Heads and Apollo Bay. While some species’ ranges might expand as a result of climate change, we might expect concomitant range contractions of other species. For example, M. australis is a primarily Tasmanian species that only reaches the mainland in the vicinity of Bass Strait. As water temperatures increase in Victoria, species such as M. australis may decline in abundance in Victoria. Declines such as this could be driven by any number of factors, including increased metabolic demand associated with warmer waters or inter-specific competition. Measuring species’ responses to climate change requires an understanding of species’ distributions, such as that provided by this research. The patterns of abundance and distribution that we have quantified provide both a baseline against which future changes may be measured and an understanding of the factors that structure populations in this region.

- 95 -

LITERATURE CITED Bernardi G (2000) Barriers to gene flow in Embiotoca jacksoni, a marine fish lacking a pelagic larval stage. Evolution 54:226-237. Booth DJ, Figueira WF, Gregson MA, Brown L, Beretta G (2007) Occurrence of tropical fishes in temperate southeastern Australia: Role of the East Australian Current. Estuarine Coastal and Shelf Science 72:102-114. Burridge CP (2000) Biogeographic history of geminate cirrhitoids (Perciformes: Cirrhitoidea) with east-west allopatric distributions across southern Australia, based on molecular data. Global Ecology and Biogeography 9:517-525. Clarke KR, Warwick RM (2001) Change in Marine Communities: An Approach to Statistical Analysis and Interpretation. (Primer-E Ltd.: Plymouth, U.K.) Colton MA, Swearer SE (2010) A comparison of two survey methods: differences between underwater visual census and baited remote underwater video. Marine Ecology-Progress Series 400:19-36. Curley BG, Gillings MR (2009) Population connectivity in the temperate damselfish Parma microlepis: analyses of genetic structure across multiple spatial scales. Marine Biology 156:381-393. Dawson MN (2005) Incipient speciation of Catostylus mosaicus (Scyphozoa, Rhizostomeae, Catostylidae), comparative phylogeography and biogeography in south-east Australia. Journal of Biogeography 32:515-533. Edgar GJ (2005) 'Australian Marine Life: The Plants and Animals of Temperate Waters.' (Reed New Holland Publishers Pty Ltd: Sydney, Australia) Eschmeyer WN, Fricke R (2009) Catalog of Fishes electronic version (9 September 2009). http://research.calacademy.org/ichthyology/catalog/fishcatmain.asp. Fandry CB (1983) Model for the 3-dimensional structure of wind-driven and tidal circulation in Bass Strait. Australian Journal of Marine and Freshwater Research 34:121-141. Figueira WF, Booth DJ (2010) Increasing ocean temperatures allow tropical fishes to survive overwinter in temperate waters. Global Change Biology 16:506-516. Gaylord B, Gaines SD (2000) Temperature or transport? Range limits in marine species mediated solely by flow. American Naturalist 155:769-789. Gibbs CF (1991) Oceanography of Bass Strait - implication for the food supply of Little Penguins Eudyptula minor. The Emu 91:395-401. Gomon MF, Bray D, Kuiter R (2008) 'Fishes of Australia's Southern Coast.' (Reed New Holland: Sydney) Hidas EZ, Costa TL, Ayre DJ, Minchinton TE (2007) Is the species composition of rocky intertidal invertebrates across a biogeographic barrier in south-eastern Australia related to their potential for dispersal? Marine and Freshwater Research 58:835-842. Horn MH, Allen LG (1978) Distributional analysis of California coastal marine fishes. Journal of Biogeography 5:23-42. Hough D, Mahon G (1994) Biophysical classification of Victoria's marine waters. In 'Towards a Marine Regionalisation for Australia'. Sydney. (Ed. J Muldoon). (Great Barrier Reef Marine Park Authority) Hutchins JB (1987) Description of a new plesiopid fish from south-western Australia with a discussion of the zoogeography of Paraplesiops. Records of the Western Australian Museum 13:231-240. Kuiter RH (2000) 'Coastal Fishes of South-eastern Australia.' (Gary Allen Pty Ltd: Sydney) Lambeck K, Chappell J (2001) Sea level change through the last glacial cycle. Science 292:679-686. Leis JM, McCormick MI (2002) The biology, behavior, and ecology of the pelagic, larval stage of coral reef fishes. In 'Coral Reef Fishes'. (Ed. PF Sale) pp. 171-199. (Academic Press: San Diego, USA) Ling SD, Johnson CR, Ridgway K, Hobday AJ, Haddon M (2009) Climate-driven range extension of a sea urchin: inferring future trends by analysis of recent population dynamics. Global Change Biology 15:719-731.

- 96 -

Lourie SA, Vincent ACJ (2004) Using biogeography to help set priorities in marine conservation. Conservation Biology 18:1004-1020. Lyne V, Last P R, Gomon MF, Scott R, Long S, Phillips A, McArdle B, Peters D, Pigot S, Kailola P (1996) Interim Marine Bioregionalisation for Australia: Towards a National System of Marine Protected Areas. pp. 73. (CSIRO Division of Fisheries and Division of Oceanography) Malcolm WB (1960) Area of distribution, and movement of the western subspecies of the Australian "salmon", Arripis trutta esper Whitley. Marine and Freshwater Research 11:282-325. Murray SN, Littler MM (1981) Biogeographical analysis of intertidal macrophyte floras of southern California. Journal of Biogeography 8:339-351. Myers AA (1997) Biogeographic barriers and the development of marine biodiversity. Estuarine Coastal and Shelf Science 44:241-248. O'Hara TD, Poore GCB (2000) Patterns of distribution for southern Australian marine echinoderms and decapods. Journal of Biogeography 27:1321-1335. Pelc RA, Warner RR, Gaines SD (2009) Geographical patterns of genetic structure in marine species with contrasting life histories. Journal of Biogeography 36:1881-1890. Pondella DJ, Gintert BE, Cobb JR, Allen LG (2005) Biogeography of the nearshore rocky- reef fishes at the southern and Baja California islands. Journal of Biogeography 32:187- 201. Ridgway KR, Condie SA (2004) The 5500-km-long boundary flow off western and southern Australia. Journal of Geophysical Research 109:C04017-04018. Ridgway KR, Dunn JR (2003) Mesoscale structure of the mean East Australian Current System and its relationship with topography. Progress in Oceanography 56:189-222. Riginos C, Nachman MW (2001) Population subdivision in marine environments: the contributions of biogeography, geographical distance and discontinuous habitat to genetic differentiation in a blennioid fish, Axoclinus nigricaudus. Molecular Ecology 10:1439- 1453. Sandery PA, Kampf J (2005) Winter-spring flushing of Bass Strait, south-eastern Australia: a numerical modelling study. Estuarine Coastal and Shelf Science 63:23-31. Sexton JP, McIntyre PJ, Angert AL, Rice KJ (2009) Evolution and ecology of species range limits. Annual Review of Ecology Evolution and Systematics 40:415-436. Ward RD, Elliott NG (2001) Genetic population structure of species in the South East Fishery of Australia. Marine and Freshwater Research 52:563-573. Waters JM (2008) Marine biogeographical disjunction in temperate Australia: historical landbridge, contemporary currents, or both? Diversity and Distributions 14:692-700. Waters JM, King TM, O'Loughlin PM, Spencer HG (2005) Phylogeographical disjunction in abundant high-dispersal littoral gastropods. Molecular Ecology 14:2789-2802. Waters JM, O'Loughlin PM, Roy MS (2004) Cladogenesis in a starfish species complex from southern Australia: evidence for vicariant speciation? Molecular Phylogenetics and Evolution 32:236-245. Waters JM, Roy MS (2003) Marine biogeography of southern Australia: phylogeographical structure in a temperate sea-star. Journal of Biogeography 30:1787-1796. Watson DL, Harvey ES, Anderson MJ, Kendrick GA (2005) A comparison of temperate reef fish assemblages recorded by three underwater stereo-video techniques. Marine Biology 148:415-425. Wethey DS (2002) Biogeography, competition, and microclimate: The barnacle Chthamalus fragilis in New England. Integrative and Comparative Biology 42:872-880. Whitely G (1932) Marine zoogeographical regions of Australia. The Australian Naturalist 8:166-167. Willis TJ, Babcock RC (2000) A baited underwater video system for the determination of relative density of carnivorous reef fish. Marine and Freshwater Research 51:755-763.

- 97 -

CHAPTER FOUR

Relationships between geographic range size, body size and abundance in temperate marine fishes

ABSTRACT Though relationships between body size, geographic range size and abundance have been quantified in numerous taxa around the globe, relatively little research has focused on fishes. As a group, fishes exhibit a wide variety of life history strategies that may not be as size-dependent as in other taxa, suggesting that fishes may not display the predicted relationships between these variables. Here we present the first analyses of these relationships for temperate marine fishes. Both ordinary least squares regression and quantile regression revealed weak correlations among variables but failed to detect relationships that were apparent at the upper and lower bounds of scatterplots. Consequently, data were categorized to allow quantification of relationships at the maxima and minima for three populations: all fishes, perciforms and non-perciforms. The relationship between body size and abundance was expected to be described by a slope of -0.75. Maxima displayed a negative relationship that was steeper than expected, while minima showed no relationship, indicating that body size placed an ultimate constraint on density. The steepness of the slope was likely driven by the difficulty of sampling diver-averse mobile predators. The relationship between geographic range size and body size was predicted to be positive, which is what we found at the lower bound, signifying that body size determined the minimum range size which a species must occupy in order to persist. However, the maxima displayed a negative relationship that was significant for non-perciform fishes, indicating that large-bodied species’ geographic range sizes were constrained, probably by size- specific differences in dispersal ability and reproductive mode. Abundance- distribution relationships were also expected to be positive, which is what we found at the lower bound, though this was only significant for non-perciforms. Contrary to predictions, regressions at maxima were negative, though only significant for perciforms, indicating that fishes with large geographic ranges were less abundant than fishes with small ranges. We suggest that the most likely explanation is the limited dispersal capacity of many large-bodied fishes. These results underscore the importance of exploring macroecological relationships at upper and lower bounds and

- 98 - the role of life history in explaining differences in these relationships among taxa, particularly between terrestrial and marine faunas.

KEYWORDS: macroecology; body size; abundance; distribution; range size; density; fish; marine; temperate; Australia; New Zealand; maximum; minimum

INTRODUCTION The relationships between body size, abundance and geographic range size exhibit remarkable concordance across numerous taxa and ecosystems (reviewed in Brown 1984, Cotgreave 1993, Brown 1995, Borregaard & Rahbek 2010). Because body size ultimately influences life history, ecology, biogeography and evolution (Brown 1984, Kelt & Brown 1998, White et al. 2007) it is often a good predictor of a species’ density and geographic range size. Basic ecological theory also predicts that as the density of a population increases, so will the spatial area that the population occupies (reviewed in Borregaard & Rahbek 2010), resulting in a positive distribution- abundance relationship.

The relationship between body size and abundance has so frequently been found to be negative that it has been described as a “fundamental relationship in ecology” (Carbone et al. 2007). Explanations of mechanisms driving this relationship are varied. In a study of mammalian herbivores, Damuth (1981) found that the negative relationship between log body size and log abundance was described by a slope of -0.75 and proposed that this relationship is driven by invariance in energy use across body sizes, with the result that abundance is ultimately constrained by metabolism. Whether this Energetic Equivalence Rule (EER) can offer a full explanation of body size-abundance relationships is unclear (reviewed in Cotgreave 1993), though there is considerable evidence to suggest that metabolism may place an ultimate constraint on the density a species can achieve (Cotgreave 1993, McGill 2008, but see Ackerman et al. 2004).

The relationship between body size and abundance, however, is not universally negative. Some studies report a positive relationship (e.g., Nee et al. 1991b), some no relationship (e.g., Kattan 1992, Pyron 1999, Jones et al. 2002) and others a negative relationship but not -0.75 (e.g., Mohr 1940, Peters & Raelson 1984). The relationship

- 99 - is also frequently described by a triangle in which species of intermediate sizes achieve the highest abundance (e.g., Brown & Maurer 1987, Blackburn et al. 1992, Warwick & Clarke 1996, Hubble 2003). Whether this is biologically meaningful, represents inadequate sampling of smaller species (Thomson et al. 1996) or is a by- product of high species richness at intermediate body sizes is unclear (Cotgreave 1993). There is certainly evidence to suggest that the strength and shape of the relationships is dependent upon the scale of the study (White et al. 2007), and that it may be influenced by phylogenetic relatedness of the study organisms (Nee et al. 1991b, Cotgreave 1993, but see Ackerman & Bellwood 2003, Cofre et al. 2007).

In addition to energetic equivalence, the negative relationship between body size and abundance has been attributed to life history traits associated with body size. In terrestrial systems, smaller bodied species tend to be r-selected and so have an inherently larger capacity for population growth than larger bodied species (Brown 1995). Body size also dictates the smallest scale at which an organism perceives its environment (Holling 1992) with differently sized species responding to the environment at different scales (Wiens 1989). Small-bodied species can better exploit fine scale environmental heterogeneity than large-bodied species, which may allow smaller species to achieve higher local densities (Brown 1984). Similarly, because the environment is fractal it can support many small species and few large organisms (Brown 1995).

Finally, the negative relationship between body size and abundance could be an artefact of sampling. Larger species tend to have larger home range sizes, meaning that their densities will be measured over larger areas than smaller species resulting in lower densities (Blackburn & Gaston 1996). While this certainly accounts for at least part of the relationship, it is not thought to offer a complete explanation (Brown 1984, Cotgreave 1993, Blackburn & Gaston 1996).

The relationship between geographic range size and abundance, also known as the distribution-abundance relationship (Borregaard & Rahbek 2010), is so ubiquitous that it too has been called a general law of ecology (Lawton 1999). It holds for a wide diversity of organisms, from vascular plants to planktonic crustacea to terrestrial vertebrates (reviewed in Brown 1984, Brown 1995, Gaston et al. 1997, Borregaard &

- 100 -

Rahbek 2010), though, as with any rule, there are exceptions (e.g., Päivinen et al. 2005, Kolb et al. 2006, Symonds & Johnson 2006). In contrast to the relationship between body size and abundance, the relationship between geographic range size and abundance is strongest among closely related or ecologically similar taxa (Brown 1995, Gaston et al. 1997). Interestingly, the relationship may differ depending on the age of the species in question, with older species exhibiting a negative abundance- distribution relationship (Johnson 1998).

Explanations for this relationship can be grouped into three categories: (1) resources, (2) metapopulation dynamics, and (3) artefacts (reviewed in Hanski et al. 1993). Within the first category, explanations include ideal free distribution theory (reviewed in Rosenzweig & Lomolino 1997) and niche breadth (Brown 1984, Gaston et al. 1997, Borregaard & Rahbek 2010), though the latter has received some criticism (Root & Cappuccino 1992, Hanski et al. 1993, Quinn et al. 1997, Pyron 1999, Tales et al. 2004, Harcourt et al. 2005, Kolb et al. 2006, Borregaard & Rahbek 2010). Metapopulation theory also offers two mechanisms by which a positive abundance- distribution relationship could arise: the carrying capacity hypothesis (Nee et al. 1991a, as reviewed in Gaston et al. 1997) and the rescue effect (Hanski 1991, Hanski et al. 1993, as reviewed in Gaston et al. 1997, Borregaard & Rahbek 2010). The third category of explanations for distribution-abundance relationships concerns sampling artefacts. Rare species are more likely to be overlooked in a sampling program than abundant species, with the consequence that their range sizes could be underestimated (Brown 1984). However, it is generally agreed that this cannot wholly explain these relationships (Brown 1984, Nee et al. 1991a, Gaston et al. 1997, Borregaard & Rahbek 2010).

The relationship between geographic range size and body size has been found to be weakly positive, based on the argument that larger bodied species must occupy larger areas in order to satisfy their larger resource requirements (Brown 1995), though as always there are exceptions (e.g., Taylor & Gotelli 1994, Harcourt et al. 2005). Some argue that the causal link between geographic range size and body size is dispersal (Reaka 1980, Hawkins et al. 2000): larger species tend to be better dispersers and better dispersers tend to have larger geographic range sizes (but see Jones et al. 2002, Lester & Ruttenberg 2005). In the marine environment, the majority of dispersal

- 101 - occurs during the larval stage and planktonic larval duration (PLD) is frequently used as a surrogate for dispersal potential (Lester & Ruttenberg 2005), though the relationship between PLD and dispersal is debated (Hawkins et al. 2000, Jones et al. 2002, Pelc et al. 2009, Weersing & Toonen 2009). As body size has been shown to be positively correlated, albeit weakly, with PLD in some fishes (Bradbury et al. 2008) we might expect a positive correlation between geographic range size and body size in marine fishes.

Explorations of relationships between body size, geographic range size and abundance have been limited to terrestrial, freshwater and intertidal systems, with little research focused on whether these relationships exist for marine fishes and if they are similar to those reported for other taxa. The few examinations of macroecological relationships in marine fishes have focused on tropical taxa and have found equivocal support for the previously reported patterns (Jones et al. 2002, Ackerman & Bellwood 2003, Ackerman et al. 2004). However, tropical and temperate species may be quite different (Peters & Raelson 1984, Mora et al. 2003), which is supported by recent research that found fishes at high latitudes to be larger and less fecund than tropical species (Fisher et al. 2010). In this research, we explore relationships between body size, geographic range size and density, used as an estimate of abundance, for the temperate marine ichthyofauna of southern Australia and New Zealand. Unlike many previous studies, we did not rely on estimates of abundance gleaned from the literature but instead measured density over a 3-year period and across a very large spatial area that encompassed most of the range of the study species.

MATERIALS & METHODS A total of 159 species of fish from 2 classes, 13 orders and 45 families were included in these analyses (Table 1). Species were included if they: (1) had a distribution limited to Australia and/or New Zealand; (2) were reported to inhabit nearshore environments including rocky reefs; and (3) had a depth range including 0-20 m and were not known to preferentially occur below 20 m. Distributional data were obtained from guidebooks (Kuiter 2000, Francis 2001, Edgar 2005, Gomon et al. 2008) and taxonomic nomenclature follows Eschmeyer & Fricke (2009).

Fish density was measured using underwater visual census (UVC) conducted by A.

- 102 -

Pérez-Matus (APM), S.D. Connell et al. (SDC), and M.A. Colton (MAC). Data were collected from varying numbers of replicates within locations nested within six regions spanning approximately 7000 km of coastline across southern Australia and the North Island of New Zealand (Table 2). Each group used SCUBA divers to identify and count all fish encountered on belt transect UVCs. SDC et al. surveyed fish along transects of width = 5 m and length = 25 m (description of sites in Connell & Irving 2009), APM used transects of width = 2 m and length = 16 m (detailed in Pérez-Matus 2010), and MAC surveyed transects of width = 5 m and mean length = 360 m (detailed in Colton & Swearer 2010). As there is evidence to suggest that transect length does not affect measures of density (e.g., Fowler 1987, Colton & Swearer 2010), species’ densities from all groups were merged in order to determine mean densities.

We estimated abundance as mean density computed for each species using all replicates from only the location(s) in which it was observed. This avoided underestimating density through the inclusion of locations in which a species did not occur. Geographic range size was measured using data in guidebooks (Kuiter 2000, Francis 2001, Gomon et al. 2008) and the GIS program DIVA (Hijmans et al. 2009). A grid of 1°x1° cells was overlaid upon a map of Australia and New Zealand. All cells touching the coastline of either country were marked, and the number of marked cells which comprised the geographic distribution of a species counted. This approach allowed for the exclusion of known gaps in species’ distributions. We used the maximum reported length (usually total length; occasionally fork or standard length) in the aforementioned guidebooks as an estimate of species’ size, with one exception: Orectolobus maculatus has a maximum reported length of 320 cm, though most individuals only reach 180 cm (Compagno 2001). Because this is such a large difference and because 320 cm was an outlier (the next biggest fish was 200 cm TL), we chose to use a total length of 180 cm for O. maculatus.

Unfortunately, the phylogeny of species included in these analyses has not been resolved and even taxonomic relationships are unclear. There is evidence to suggest that some macroecological relationships are influenced by phylogenetic relatedness (Cotgreave 1993, Borregaard & Rahbek 2010), which makes data points non- independent. To control for this, we separated the species into two groups:

- 103 -

Table 1 (This and following pages): Species included in analyses by family and observer, with APM = A. Pérez-Matus, MAC = M. Colton, and SDC = S. Connell. Family Species APM MAC SDC Acanthuridae Prionurus microlepidotus 

Aplodactylidae Aplodactylus arctidens  

Aplodactylidae Aplodactylus lophodon  

Aplodactylidae Aplodactylus westralis 

Apogonidae Siphamia cephalotes 

Aracanidae Aracana aurita    Aracanidae Aracana ornata  

Arripididae Arripis georgianus 

Arripididae Arripis trutta 

Aulopidae Aulopus purpurissatus  

Bovichtidae Bovichtus angustifrons 

Carangidae Decapterus koheru 

Carangidae Trachurus novaezelandiae 

Chaetodontidae Chelmonops curiosus 

Chaetodontidae Chelmonops truncatus 

Cheilodactylidae Cheilodactylus fuscus  

Cheilodactylidae Cheilodactylus gibbosus 

Cheilodactylidae Cheilodactylus nigripes    Cheilodactylidae Cheilodactylus rubrolabiatus 

Cheilodactylidae Cheilodactylus spectabilis    Cheilodactylidae Dactylophora nigricans  

Cheilodactylidae Nemadactylus douglasii    Chironemidae Chironemus marmoratus  

Clinidae Heteroclinus johnstoni 

Congridae Conger verreauxi 

Dinolestidae Dinolestes leweni    Diodontidae Dicotylichthys punctulatus  

Diodontidae Diodon nicthemerus    Enoplosidae Enoplosus armatus  

Gerreidae Parequula melbournensis 

Gobiesocidae Cochleoceps orientalis 

Heterdontidae Heterodontus portusjacksoni  

Kyphosidae Atypichthys strigatus  

Kyphosidae Girella elevata 

Kyphosidae Girella tephraeops 

Kyphosidae Girella tricuspidata  

Kyphosidae Girella zebra  

Kyphosidae Kyphosus sydneyanus  

Kyphosidae Scorpis aequipinnis  

Kyphosidae Scorpis georgiana 

Kyphosidae Scorpis lineolata    Kyphosidae Scorpis violaceus 

Kyphosidae Tilodon sexfasciatum  

- 104 -

Table 1 Continued Family Species APM MAC SDC Labridae Achoerodus gouldii 

Labridae Achoerodus viridis  

Labridae Austrolabrus maculatus 

Labridae Coris auricularis 

Labridae Coris picta  

Labridae Coris sandeyeri  

Labridae Dotalabrus alleni 

Labridae Dotalabrus aurantiacus    Labridae Eupetrichthys angustipes    Labridae Halichoeres brownfieldi 

Labridae Notolabrus celidotus 

Labridae Notolabrus fucicola    Labridae Notolabrus gymnogenis  

Labridae Notolabrus inscriptus 

Labridae Notolabrus parilus 

Labridae Notolabrus tetricus    Labridae Ophthalmolepis lineolata  

Labridae Pictilabrus laticlavius    Labridae Pictilabrus viridis 

Labridae Pseudolabrus biserialis 

Labridae Pseudolabrus luculentus  

Labridae Pseudolabrus miles 

Labridae Pseudolabrus psittaculus    Latrididae Latridopsis ciliaris 

Latrididae Latridopsis forsteri    Monacanthidae Acanthaluteres brownii 

Monacanthidae Acanthaluteres spilomelanurus 

Monacanthidae Acanthaluteres vittiger    Monacanthidae Brachaluteres jacksonianus 

Monacanthidae Eubalichthys bucephalus  

Monacanthidae Eubalichthys mosaicus 

Monacanthidae Meuschenia australis  

Monacanthidae Meuschenia flavolineata    Monacanthidae Meuschenia freycineti    Monacanthidae Meuschenia galii  

Monacanthidae Meuschenia hippocrepis  

Monacanthidae Meuschenia scaber  

Monacanthidae Meuschenia trachylepis 

Monacanthidae Scobinichthys granulatus  

Monodactylidae Schuettea woodwardi 

Moridae Lotella rhacina    Moridae Pseudophycis barbata 

Mullidae Upeneichthys lineatus  

Mullidae Upeneichthys vlamingii 

- 105 -

Table 1 Continued Family Species APM MAC SDC Mullidae Upeneus francisi 

Muraenidae Gymnothorax prasinus    Myliobatidae Myliobatis australis 

Myliobatidae Myliobatis tenuicaudatus 

Odacidae Haletta semifasciata 

Odacidae Neoodax balteatus    Odacidae Odax acroptilus    Odacidae Odax cyanomelas    Odacidae Odax pullus 

Odacidae Siphonognathus attenuatus  

Odacidae Siphonognathus beddomei  

Odacidae Siphonognathus caninus 

Orectolobidae Orectolobus maculatus 

Parascylliidae Parascyllium variolatum 

Pempherididae Pempheris adspersa 

Pempherididae Pempheris affinis    Pempherididae Pempheris analis 

Pempherididae Pempheris compressa 

Pempherididae Pempheris klunzingeri 

Pempherididae Pempheris multiradiata  

Pentacerotidae Pentaceropsis recurvirostris    Pinguipedidae Parapercis collias 

Plesiopidae Paraplesiops meleagris 

Plesiopidae Trachinops brauni 

Plesiopidae Trachinops caudimaculatus  

Plesiopidae Trachinops noarlungae  

Plesiopidae Trachinops taeniatus  

Pomacentridae Chromis dispilus 

Pomacentridae Chromis hypsilepis  

Pomacentridae Chromis klunzingeri 

Pomacentridae Chromis westaustralis 

Pomacentridae Mecaenichthys immaculatus 

Pomacentridae Parma alboscapularis 

Pomacentridae Parma mccullochi 

Pomacentridae Parma microlepis  

Pomacentridae Parma unifasciata 

Pomacentridae Parma victoriae  

Scorpaenidae Scorpaena cardinalis  

Scorpaenidae Scorpaena papillosa 

Scyliorhinidae Cephaloscyllium laticeps  

Serranidae Acanthistius ocellatus 

Serranidae Caesioperca lepidoptera  

Serranidae Caesioperca rasor    Serranidae Epinephelides armatus 

- 106 -

Table 1 Continued Family Species APM MAC SDC Serranidae Hypoplectrodes huntii 

Serranidae Hypoplectrodes maccullochi    Serranidae Hypoplectrodes nigroruber 

Serranidae Othos dentex 

Sillaginidae Sillaginodes punctatus 

Sparidae Acanthopagrus australis 

Sparidae Chrysophrys auratus  

Sphyraenidae Sphyraena novaehollandiae 

Syngnathidae Phycodurus eques 

Syngnathidae Phyllopteryx taeniolatus  

Tetraodontidae Canthigaster callisterna 

Tetraodontidae Omegophora cyanopunctata 

Tetraodontidae Tetractenos glaber  

Trachichthyidae Optivus agastos  

Trachichthyidae Optivus elongatus 

Trachichthyidae Trachichthys australis  

Tripterygiidae Forsterygion flavonigrum 

Tripterygiidae Forsterygion lapillum 

Tripterygiidae Forsterygion malcolmi 

Tripterygiidae Forsterygion maryannae 

Tripterygiidae Forsterygion varium 

Tripterygiidae Karalepis stewarti 

Tripterygiidae Notoclinops caerulepunctus 

Tripterygiidae Notoclinops segmentatus 

Tripterygiidae Notoclinops yaldwyni 

Tripterygiidae Ruanoho decemdigitatus 

Tripterygiidae Ruanoho whero 

Urolophidae Urolophus cruciatus  

Perciformes and non-Perciformes, though both groups are polyphyletic/paraphyletic (Smith & Wheeler 2006), which is an approach similar to that used in other studies lacking resolved phylogenies (e.g., Nee et al. 1991b, Lester & Ruttenberg 2005). All analyses were conducted on these two taxa as well as on all fishes.

Ordinary least squares & quantile regressions

Regressions were fit to loge-transformed data in R (v.2.9.2) using the package LMODEL2 for ordinary least squares (OLS) regression, and the package

QUANTREG for quantile regression (QR). A simple linear model, yi = β0 + β1xi + εi was compared to a null model, yi = 1 + εi. For each model, seven quantiles (τ) were examined: 0.05, 0.25, 0.5, 0.75, 0.90, and 0.95. For each of these, SAF, the weighted

- 107 - sum of absolute deviations minimized in estimating the τth quantile (Cade et al. 2005), was computed using source code from Lancaster and Belyea (2006). Model fit for OLS was assessed using r2 values, and the null model was compared to the full model using F-ratios computed in SPSS. Model fit for QR was assessed using R1 values, 2 -1 which are analogous to r values, computed as (SAFnull – SAFmodel)(SAFnull) . Also using code from Lancaster and Belyea (2006), AICc (Akaike Information Criterion) values were computed for each level of τ. AICc values were compared between the full and null models, and Δi computed as the absolute value of the difference between the AICc,full and AICc,null. A smaller AICc value indicates a model that better describes the data, and Δi values ≥ 2 indicated support for the full model (Lancaster & Belyea 2006).

Relationships at upper and lower bounds Scatterplots of the variables (Fig. 1) suggested the existence of relationships at maxima and minima that were not resolved by either OLS or QR (see Results). To explore these, we categorized x-axis variables and determined the maxima and minima for each of the bins (sensu Blackburn et al. 1992). The choice of bin number determines the strength and perhaps the shape of relationships; if too few bins are used, the correlation between variables will be misleadingly high, and if too many bins are used the regressions will no longer fit the upper or lower bounds (Blackburn et al. 1992). Most approaches to the selection of bin number, however, have been developed to normalize histograms (e.g., Sturges 1926, Scott 1979, Taylor 1987). Here, we applied the approach of Blackburn et al. (1992) by fitting relationships between variables using a variety of bin numbers, and plotting slopes and correlations as a function of bin number. As with the data of Blackburn et al. (1992), we found that slopes changed little when between approximately 4 and 20 bins were used (Fig. 2a,c,e), though the associated correlation strength changed considerably (Fig. 2b,d,f).

Though the choice of bin numbers from a set of reasonable bin numbers is arbitrary (Blackburn et al. 1992), we developed several criteria to aid with bin number selection. For each relationship, we selected the largest number of bins in which (1) each category contained≥ 1 datum, and (2) no category had maximum = minimum (i.e., n = 1). However, when the population was reduced to perciform and non- perciform fishes, satisfying these criteria was not possible with more than 5 bins. In

- 108 -

Table 2: The location and amount of sampling as conducted by each of three research groups: SDC = Sean Connell et al.; MAC = Madhavi Colton; APM = Alejandro Pérez-Matus. Region abbreviations are WA = Western Australia, SA = South Australia, VIC = Victoria, NSW = New South Wales, TAS = Tasmania and NZ = New Zealand. A replicate for SDC is a UVC belt transect of width = 5 m and length = 25 m; a replicate for MAC is a UVC transect of width = 5 m and mean length = 360 m; and a replicate for APM is a belt transect of width = 2 m and length = 16 m. Research group Region Locations # Replicates SDC WA Cape Leeuwin 8 Albany 8 Bremer 8 Esperance 8 SDC SA Elliston 8 Pt. Lincoln 8 West Cape 8 Fleurieu Peninsula 8 MAC VIC Apollo Bay 12 -

109 Barwon Heads 7

Wilsons Promontory west 8 - Wilsons Promontory east 6 Cape Conran 6 Cape Howe 8 SDC NSW Eden 8 Batemans Bay 8 Jervis Bay 8 Sydney 8 APM TAS Maria I. 150 Bruny I., Betsey I., Southport 149 APM NZ Auckland 150 Kapiti I. 150 Poor Knights Islands 150 Wellington 150

- 109 -

these cases, the largest number of bins that contained ≤ 1 empty category and ≤ 2 bins with n = 1 was chosen. Within each bin, the maximum and minimum density or geographic range size (depending on the relationship) was identified and then loge- transformed. Linear regressions were fit to relationships at the upper and lower bounds using SPSS (v.16.0). T-tests were used to test the null hypothesis of no relationship between the variables, i.e., slope = zero.

RESULTS Body size ranged from 3.6 cm for Notoclinops caerulepunctus (Tripterygiidae) to 200 cm for Conger verreauxi (Congridae). Geographic range size varied from 9 quadrats for Pempheris adspersa (Pempheridae) to 189 quadrats for Chrysophrys auratus (Sparidae). The least abundant species was Haletta semifasciata (Odacidae) with a density of 0.00004 m-2, and the most abundant was Chromis dispilus (Pomacentridae) with a density of 0.16 m-2.

Ordinary least squares & quantile regressions The highest regression coefficient for OLS regressions was r2 = 0.08 for a positive relationship between geographic range size and density using data from all species (Table 3). The relationships between body size and abundance and between abundance and geographic range size were both negative and extremely weak. The relationships for perciforms and non-perciforms were no stronger (not shown).

For QR, the highest regression coefficient was R1 = 0.1 for the distribution-abundance relationship at τ = 0.05 (Table 3). All other relationships were≤ 0.07, and for only some relationships did Δi indicate support for the full model (Table 3).

Relationships at upper and lower bounds We found strong negative relationships, with slopes significantly different to zero, between body size and maximum abundance for all groups of fishes (Fig. 3a,b,c). Phylogeny, in the way that we controlled for it, did not affect the slope of the line though the intercept was smaller for non-perciforms than for either of the other two populations. The relationship was strongest when all fishes were included, and non- perciforms had the least steep slope (Fig. 3). When 101 bins were included in the relationship, the slope was -0.97 (Fig. 2a), which is closest to the value of -0.75

- 110 -

Table 3: Results from ordinary least squares regression (OLS) and quantile regression (QR). “Dens” = density, “lmax” = total length, and “geog” = 1 2 geographic range size. R is analogous to r in OLS. For QR, significant intercepts (β0) and slopes (β1) are in bold, with significance determined using a t-test at α = 0.05. Δi is the absolute value of the difference between the full model AICc and the null model AICc, with Δi ≥ 2 indicating support for the full model (values in bold). Population is all fishes. 2 1 Model Model τ r or R F-ratio p β0 ± SE β1 ± SE AICc Δi . OLS ln(dens) = β0 + β1 ln(lmax) 0.04 6.612 0.011 -4.21 ± 0.56 -0.41 ± 0.16 . OLS ln(dens) = β0 + β1 ln(geog) 0.03 4.717 0.03 -3.72 ± 0.87 -0.47 ± 0.22 . OLS ln(geog) = β0 + β1 ln(lmax) 0.08 13.75 0.000 3.28 ± 0.20 0.21 ± 0.06 . QR ln(dens) = β0 + β1 ln(lmax) 0.05 0.00 -8.58 ± 1.28 -0.03 ± 0.37 729.81 1.20 0.10 0.01 -6.25 ± 1.03 -0.44 ± 0.23 700.67 1.41 0.25 0.03 -4.99 ± 0.80 -0.48 ± 0.23 646.52 6.03 0.50 0.03 -4.41 ± 0.73 -0.39 ± 0.21 634.58 7.20 0.75 0.01 -3.14 ± 0.88 -0.36 ± 0.24 657.47 1.26 0.90 0.02 -1.99 ± 1.03 -0.41 ± 0.29 691.45 4.11

- 0.95 0.03 -0.88 ± 0.87 -0.52 ± 0.21 712.12 8.89

111 . QR ln(dens) = β0 + β1 ln(geog) 0.05 0.00 -8.69 ± 0.134 0.00 ± 0.38 730.66 2.05

0.10 0.00 -8.04 ± 2.14 0.06 ± 0.51 703.73 1.65 - 0.25 0.00 -6.25 ± 1.63 -0.11 ± 0.38 653.88 1.33 0.50 0.02 -2.90 ± 1.12 -0.70 ± 0.28 636.05 5.73 0.75 0.02 -2.45 ± 1.29 -0.18 ± 0.31 653.26 5.47 0.90 0.02 -1.70 ± 1.67 -0.47 ± 0.41 692.33 3.23 0.95 0.03 -0.20 ± 1.94 -0.61 ± 0.46 712.53 8.49 . QR ln(geog) = β0 + β1 ln(lmax) 0.05 0.10 1.60 ± 0.21 0.39 ± 0.04 393.27 32.62 0.10 0.07 2.24 ± 0.25 0.26 ± 0.06 381.33 19.71 0.25 0.04 2.47 ± 0.43 0.31 ± 0.12 359.67 10.92 0.50 0.03 3.74 ± 0.27 0.11 ± 0.08 311.40 8.37 0.75 0.04 3.78 ± 0.18 0.20 ± 0.05 283.97 11.17 0.90 0.03 3.90 ± 0.24 0.24 ± 0.06 302.43 9.02 0.95 0.07 4.33 ± 0.25 0.14 ± 0.08 321.90 21.64

- 111 -

predicted by EER. Relationships between body size and abundance at the lower bound for all populations of fishes were very weak and slopes were not significantly different to zero (Fig. 3a,b,c).

Between range size and body size, we found a positive relationship with slope values significantly different to zero for minima (Fig. 3d,e,f). In contrast, only one of the relationships for maximum range size, that for non-perciform fishes, had a slope significantly different to zero at p = 0.050, and this relationship was strongly negative (Fig. 3d,e,f). Given the low number of bins used in constructing regressions at upper and lower bounds (i.e., the tests had low power), this relationship is most likely significant.

We found a positive relationship between minimum density and geographic range size for all three populations, though the slope was marginally insignificantly different from zero (Fig. 3g,h,i). The relationship was strongest for non-perciform fishes. In contrast, maximum density was negatively related to geographic range size (Fig. 3g,h,i), but the slope was only significantly different from zero for perciform fishes.

DISCUSSION This research represents the first time that macroecological relationships between body size, geographic range size and abundance have been explored for temperate marine fishes. Though these relationships were highly variable, they showed evidence of constraints wherein body size constrained abundance and geographic range size, and range size limited abundance. It is well recognized that a species’ abundance and distribution are the result of many processes (e.g., behaviour, evolution, larval dispersal) acting upon populations (Mora et al. 2003 and references therein). The variability in macroecological relationships, including those described here, may be attributed to the multitude of ecological processes that shape population parameters beyond upper and lower bounds (Brown 1995).

Body size and abundance The strong negative relationship between body size and abundance at the upper bound indicates that abundance of large-bodied temperate fishes is constrained, while the lack of a relationship at the lower bound suggests that both small- and large-bodied

- 112 -

a) 0.18 b) 0 0.16 -2 0.14 0.12 -4 0.1 -6 0.08 (Density) Density

0.06 ln -8 0.04 -10 0.02 0 -12 0 100 200 300 400 0 2 4 6 8 Body size ln(Body size) c) 200 d) 6 180 160 5 140 4 120 100 3 80 60 2 40 (Geographic range size) range (Geographic Geographic range size range Geographic 1

20 ln 0 0 0 100 200 300 400 0 2 4 6 8 Body size ln(Body size) e) 0.18 f) 0 0.16 -2 0.14 0.12 -4 0.1 -6 0.08 (Density) Density

0.06 ln -8 0.04 -10 0.02 0 -12 0 50 100 150 200 0 2 4 6 Geographic range size ln(Geographic range size) Figure 1: Scatter plots of relationships between (a,c,e) raw and (b,d,f) transformed variables. Body size is maximum reported length for the species, geographic range size is measured in quadrats, and density is m-2. Population is all fishes.

organisms can occur at low densities (Fig. 3a,b,c). This contrasts with the results of Jones et al. (2002) who found body size to be a poor predictor of abundance in coral reef fishes. These authors, however, modelled regressions at the mean, which may have masked constraining relationships (Blackburn et al. 1992). Similarly, Hubble (2003) modelled the relationship between body size and abundance for coral reef labrids and found evidence of a triangular relationship in which species of

- 113 -

intermediate size were the most abundant (also see Cotgreave 1993, Brown 1995, White et al. 2007). A triangular relationship was also reported by Ackerman et al. (2004) who found that the relationship between mass and density for small reef fishes was different to that of large reef fishes; the larger species showed a negative relationship consistent with the energetic equivalence rule (EER), while the smallest species did not. The authors proposed that the smallest fishes may access resources in a different way to larger species – for example, much of the somatic growth of small- bodied species occurs in the plankton as opposed to on the reef, which is the case for larger-bodied species. In this research, a scatterplot of loge-transformed data shows evidence of a triangular relationship (Fig. 1b), though this was largely removed by the binning procedure (Fig. 3a,b,c).

Triangular or polygonal relationships have been reported in the literature and a variety of explanations offered for their existence (reviewed in Blackburn & Gaston 1997). A triangular relationship could be a statistical artefact driven by high species richness of intermediate sized organisms (Cotgreave 1993). However, in this research the most speciose group was small-bodied (Fig. 4), indicating a lack of support for this explanation. Triangular relationships have also been attributed to under-sampling of small species (Thomson et al. 1996). UVC is well known to be poor at sampling small cryptic species (Willis 2001, Watson et al. 2005), which could result in truncation of these relationships and low densities for small-bodied species. In a study that used the ichthyocide rotenone to ensure complete sampling of the smallest species of tropical reef fishes, Ackerman & Bellwood (2003) found that mass was negatively related to density, and that the relationship was strongest at upper and lower bounds. They argued that truncation may indeed account for the polygonal and triangular relationships reported in the literature. In this research, several of the smallest species observed form dense schools over reefs, e.g., Trachinops taeniatus (Plesiopidae) and Chromis hypsilepis (Pomacentridae), and so were unlikely to be overlooked by divers. However, these species are unlikely to be the smallest species in the locations sampled by this research. For example, many species in the family Tripterygiidae and the genus Heteroclinus are very small (<10 cm TL) and cryptic. The lack of complete sampling of these types of species could result in the triangular relationship we observed, and the weakness of the triangular relationship could be due to the fact that several smaller species occur at high densities (Fig. 3b).

- 114 -

a) 15 Max b) 1.2 Max Min Min 10 1.0 0.8 5 0.6 0 Correlation Slope 0.4 0 20 40 60 80 100 -5 0.2

-10 0.0 0 20 40 60 80 100 120 -15 # Bins # Bins -

115 Figure 2 (This and following page): Slope (a,c,e) and correlation strength (R2) (b,d,f) for upper (max) and lower (min) data as a function of the

number of bins for (a,b) lmax and density, (c,d) lmax and geog, and (e,f) geog and density. 95% confidence intervals in (a,c,e) are indicated by - dashed lines for maxima and dotted lines for minima. Arrows pointing to x-axes indicate the number of bins selected for analyses. The arrow on the y-axis in (a) indicates the expected slope based on the principle of energetic equivalence. Population is all fishes and data are log-log transformed.

- 115 -

c) 4 Max d) 1.2 Max Min Min 3 1.0 2 0.8 1 0.6 0 Correlation Slope 0.4 0 20 40 60 80 100 -1 0.2 -2 0.0 -3 0 20 40 60 80 100 120 -4 # Bins # Bins -

116

Max - e) 30 f) 1.2 Max Min Min 20 1.0

0.8 10 0.6 0 Correlation Slope 0 20 40 60 80 0.4 -10 0.2

-20 0.0 0 20 40 60 80 100 -30 # Bins # Bins

Figure 2 Continued - 116 -

The relationships between body size and abundance (Fig. 3a,b,c) did not have a slope close to that predicted by EER (Damuth 1981), and were instead much steeper than has been reported in the literature (reviewed in Blackburn et al. 1992, White et al. 2007). Carbone et al. (2007) were able to generate relationships between body size and abundance that were more steeply negative than that predicted by EER by modelling species inhabiting three-dimensional environments in which prey were clumped. However, the slopes reported in that study were not as steep as the slopes we reported here, which indicate that small-bodied fishes are more dense and/or large- bodied fishes are less dense than other taxa. Both of these explanations could be valid.

Two of the three most abundant species we observed are small-bodied and form dense schools above the benthos. Because the habitat they use is three dimensional, such species may be able to achieve much higher density than many intertidal or terrestrial taxa. Alternatively, an explanation can be proposed for the lower density of large fishes. The largest species in this research were all mobile predators, which are notoriously under-sampled by UVC (Willis & Babcock 2000, Watson et al. 2005, Colton & Swearer 2010). This could result in an underestimate of the density of the largest fishes and the observed very steep slope relating body size to density. Then again, larger species could simply be more rare in marine than terrestrial systems, which could either be a natural state or the result of fishing, which selectively targets large individuals and species resulting in the observed global decline in large fishes (Myers & Worm 2003).

The relationship between body size and abundance had a similar slope for each of our taxonomic groupings though the intercept was lowest for non-perciforms (Fig. 3a,b,c). In contrast, Nee et al. (1991b) found that the relationship between body size and abundance in British birds was negative across all species, non-existent within passerines and non-passerines, and positive when examined within a tribe. However, Cofre et al. (2007) found that phylogenetic relatedness did not affect the relationship between abundance and body size in Chilean birds, and Ackerman & Bellwood (2003) found no measureable effect of phylogeny on this relationship in tropical fishes. It is possible that our own data would reveal phylogenetic effects on the body size-abundance relationship were we able to analyse them using independent contrasts.

- 117 -

Body size and geographic range size The relationship between body size and geographic range size was predicted to be weakly positive, largely driven by the fact that larger-bodied species have relatively larger resource requirements that necessitate their occupation of larger areas (Brown 1995, Gaston et al. 1997). We found a strong positive relationship at the lower bound between range size and body size in each population (Fig. 3d,e,f), suggesting that no matter what size a species achieves, there is a minimum range which it must occupy most likely in order to fulfil its resource requirements. This differs from the results of Lester & Ruttenberg (2005) who reported that body size and range size were only weakly correlated in tropical Indo-Pacific fishes. However, these correlations were explored at the mean which could conceal constraining relationships (Blackburn et al. 1992). In contrast, we found non-perciforms to have a strong negative relationship between body size and range size at the upper bound, indicating that small-bodied non-perciform fishes were more widespread than those of larger sizes (Fig. 3f). We suggest that this may be driven by life-history traits, in particular reproductive mode. In many marine organisms, the majority of dispersal occurs during early life history stages, and there is some evidence to suggest that pelagic larval duration (PLD) may be linked to range size (Hawkins et al. 2000, Mora et al. 2003, Lester & Ruttenberg 2005, Pelc et al. 2009, but see Jones et al. 2002, Weersing & Toonen 2009). Four of the six largest species observed in this study were elasmobranchs, which either lay benthic eggs (Heterodontus portusjacksoni) or are ovoviviparous (Myliobatis australis, M. tenuicaudatus, and Orectolobus ornatus) (Compagno 2001) and thus have limited potential for dispersal during their early life stages compared to other, small-bodied fishes (Bradbury et al. 2008). If larger-bodied marine fishes have limited dispersal relative to small-bodied fishes, we would predict they would have smaller range sizes, which is what we observed at the upper bound (Fig. 3f). However, this contrasts with results reported by Bradbury et al. (2008) who found a positive relationship between body size and PLD in a meta-analysis of global fishes. However, as the authors noted, the dominance in their data of low-dispersal taxa from low latitudes may make extrapolation of their results to the highly dispersive taxa inhabiting high latitudes problematic. In this research, it certainly seems likely that the prevalence of large-bodied elasmobranchs with low capacity for dispersal could account for the negative relationship between range size and body size.

- 118 -

Max a) 0 ymax = -1.91x + 4.20 R² = 0.76 Min

-2 tslope = -4.7, p = 0.002

-4

-6 (Density)

ln -8

ymin = 0.72x - 11.20 -10 R² = 0.21 t = 1.4, p = 0.2 -12 slope 0.00 1.00 2.00 3.00 4.00 5.00 6.00 ln(Body size bin mid-point)

b) 0 ymax = -1.56x + 2.99 Max R² = 0.71

-2 tslope = -3.5, p = 0.02

-4

-6 (Density)

ln -8 ymin = 1.01x - 12.17 R² = 0.31 -10 t slope= 1.5, p = 0.2

-12 0 1 2 3 4 5 6 ln(Body size bin mid-point)

c) 0 Max Min

-2

ymax = -1.42x + 0.12 -4 R² = 0.63

tslope = -2.9, p = 0.03

(Density) -6 ln y = 0.08x - 8.65 -8 min R² = 0.01

tslope = 0.2, p = 0.8 -10 0 1 2 3 4 5 6 ln(Body size bin mid-point)

Figure 3 (This and following pages): Relationships at upper and lower bounds for (a-c) body size (cm) and density (m-2), (d-f) body size and geographic range size (# quadrats), and (g-i) geographic range size (geog) and density (dens). Data are log-log transformed and the population is (a,d,g) all fishes, (b,e,h) perciform fishes, and (c,f,i) non-perciform fishes. The x-axis is the natural log of the mid-bin value. Results and significance of t=tests testing the H0: slope = 0.

- 119 -

Max d) 6 ymax = -0.11x + 5.31 R² = 0.09 Min 5 tslope = -0.8, p = 0.4

3 y = 0.81x + 0.03 2 min R² = 0.86

(Geographic range size) range (Geographic t slope= 6.4, p < 0.0005 ln 1

0 0 1 2 3 4 5 6 ln(Body size bin mid-point)

Max e) 6 ymax = -0.14x + 5.37 R² = 0.08 Min 5 tslope = -0.7, p = 0.5

ymin = 0.73x + 0.29 2 R² = 0.89

(Geographic range size) range (Geographic t slope= 6.3, p = 0.001

ln 1

0 0 1 2 3 4 5 6 ln(Body size bin mid-point)

Max f) 6 ymax = -0.31x + 6.04 R² = 0.57 Min 5 t slope= 2.6, p = 0.05

ymin = 0.73x + 0.68 2 R² = 0.78

t slope= 4.2, p = 0.009 (Geographic range size) range (Geographic

ln 1

0 0 1 2 3 4 5 6 ln(Body size bin mid-point)

Figure 3 Continued

- 120 -

g) 0 ymax = -1.07x + 1.74 R² = 0.48 -2 t slope = -2.1, p = 0.09

-4

-6 (Density)

ln -8

ymin = 1.32x - 13.74 -10 R² = 0.27

t slope= 1.4, p = 0.2 -12 0.00 1.00 2.00 3.00 4.00 5.00 6.00 ln(Geographic range size bin mid-point)

0 Max h) ymax = -1.38x + 2.95 R² = 0.63 Min -2 tslope = -2.9, p = 0.03

-4

-6 (Density) ln -8 ymin = 1.41x - 14.25 R² = 0.33 -10 tslope = 1.6, p = 0.2

-12 0 1 2 3 4 5 6 ln(Geographic range size bin mid-point) i) 0 Max Min

-2 ymax = -0.60x - 2.59 R² = 0.18 t = -1.0, p = 0.4 -4 slope

-6 (Density) ln -8

y = 1.93x - 16.02 -10 min R² = 0.53

tslope = -2.4, p = 0.06 -12 0 1 2 3 4 5 6 ln(Geographic range size bin mid-point)

Figure 3 Continued

- 121 -

Geographic range size and abundance Theory predicts that the relationship between geographic range size and abundance will be positive, based on Ideal Free Distribution Theory (Fretwell & Lucas Jr. 1969, Gaston et al. 1997), metapopulation dynamics (Hanski et al. 1993, Gaston et al. 1997), and a variety of resource-based hypotheses (e.g., niche breadth, resource availability) (Brown 1984, Borregaard & Rahbek 2010). Although all populations showed a positive distribution-abundance relationship at the lower bound, only non- perciform fishes had a strong relationship, though the slope was not significantly different to zero at p = 0.06 (Fig. 3g,h,i). These data suggest that for a given density there is a minimum range size that a species must occupy, which is likely driven by a species’ need for resources, including space. However, these data also indicated that geographic range size was a poor predictor of density in temperate marine fishes, which is consistent with the findings of Jones et al. (2002).

Interestingly, the distribution-abundance relationship showed much greater variation in density of restricted species than in the density of widespread species (Fig. 1e). This could be because there were few widespread species in the present study (Fig. 4) and so by chance alone we would expect to find the highest and lowest abundance values in species with restricted ranges. This is similar to the explanation offered for the triangular relationship between body size and abundance (Cotgreave 1993), in which the higher abundance of species of intermediate sizes is attributed to the higher species richness of medium-sized species.

Contrary to predictions, maximum density was negatively related to geographic range size (Fig. 3g,h,i), but the slope was significantly different from zero only for perciform fishes (Fig. 3h). In other words, some perciforms occurred at high densities despite having a small geographic range size. Finnish butterflies also exhibited a negative relationship between density and distribution, which was thought to be driven by range location effects resulting from measuring density at the edge of the species’ ranges (Päivinen et al. 2005). This explanation does not apply to the results reported here as density measurements were made across the range of most species. In a particularly elegant study, Johnson (1998) demonstrated a negative distribution- abundance relationship for ancient species of Australian marsupials, and a positive relationship for species that had recently evolved. In contrast, Bellwood & Meyer

- 122 -

45 lmax 40 geog 35 30 25 20 15

Number of of Number species 10 5 0 10 70 80 90 20 30 40 50 60 - - 100 110 120 130 140 150 160 170 180 190 200 0 61 71 - 81 - 11 - 21 - 31 - 41 - 51 - 91 - 101 - 111 - 121 - 131 - 141 - 151 - 161 - 171 - 181 - 191 -

Range of body size or geographic range size Figure 4: Number of species per range of total length (lmax) (cm) or geographic range size (geog) (quadrats).

(2009) found little evidence linking geographic range size and species’ ages in tropical fishes. Although this is an intriguing avenue of exploration, without a molecular phylogeny we were unable to explore whether species’ ages can explain the negative abundance-distribution relationship that we observed. Alternatively, this negative relationship could be driven by schooling behaviour, sampling effects, or dispersal ability.

Of those species with geographic ranges in the lowest quartile, those with the highest densities were schooling species, i.e., Atypichthys strigatus (Kyphosidae), Chromis dispilus (Pomacentridae), C. klunzingeri, and Trachinops taeniatus (Plesiopodae). Schooling behaviour could allow a species to achieve high densities without occupying large areas, in a similar vein to which species inhabiting three-dimensional environments have been shown to exhibit a steep negative relationship between body size and abundance (Carbone et al. 2007). When schooling species were removed, as occurred in the analysis of non-perciform fishes, the negative relationship was weakened (Fig. 3i). However, the link between schooling behaviour and geographic range size, as opposed to area of occupancy, seems tenuous. PLD may be an important determinant of range size (Hawkins et al. 2000, Pelc et al. 2009), and there is little evidence to suggest that schooling species share a common reproductive mode or, presumably, a similar PLD. For example, T. taeniatus lays benthic eggs that are guarded by males, whereas A. strigatus has pelagic eggs (Neira et al. 1998). Without a

- 123 -

common life history, it seems unlikely that schooling species would necessarily exhibit similar geographic range sizes or densities.

As has already been discussed, UVC is especially poor at recording large-bodied mobile predators (Willis & Babcock 2000, Watson et al. 2005, Colton & Swearer 2010). Some of these species also have the largest range sizes, e.g., Chrysophrys auratus (Sparidae) and Pseudophycis barbata (Moridae). Underestimating the density of widespread mobile predators could lead to a negative abundance-distribution relationship. However, many of the widespread species recorded in this research are relatively small-bodied, e.g., Meuschenia scaber (Monacanthidae) and Caesioperca rasor (Serranidae), suggesting that under-sampling of mobile predators cannot wholly explain the negative abundance-distribution relationship we observed.

Geographic range size may ultimately be the consequence of dispersal, with better dispersers having larger ranges than those with poorer dispersal capacity (Hawkins et al. 2000, Pelc et al. 2009, but see Weersing & Toonen 2009). If, in marine systems, dispersal is greater in widespread species, populations may be chronically under- saturated leading to a slightly negative relationship between density and range size. However, Lester & Ruttenberg (2005) suggested that speciation in remote locations could result in narrow endemics evolving from species with high dispersal capacity, with the result that widely dispersing species have both large and small geographic ranges. This, they suggested, could obscure a positive distribution-abundance relationship. Of the species included in these analyses, those with the most limited dispersal were primarily non-perciforms. That we see the only significantly negative distribution-abundance relationship for perciforms supports the idea that widespread dispersal in marine communities may lead to diffuse populations. However, whether there exists a relationship between PLD and dispersal remains a matter of some debate (Hawkins et al. 2000, Jones et al. 2002, Pelc et al. 2009, Weersing & Toonen 2009). To more completely understand whether dispersal capacity influences macroecological relationships in temperate marine fishes requires a basic understanding of PLD, data which do not exist for the species in this study.

CONCLUSION Relationships between body size, geographic range size and abundance have been

- 124 -

explored in numerous taxa and, despite the remarkable concordance shown, there is on-going debate about the mechanisms driving these relationships and whether they apply to all systems. Relationships between these variables are, at least in part, thought to be driven by size-specific life history traits, including dispersal. In this research, we found that fishes conformed to predictions in only some of the relationships, showed no evidence of others, and, in some cases, defied expectations. In terrestrial systems, large-bodied species generally have a larger capacity for dispersal than small bodied species. In the marine environment, however, the situation may be the opposite. For example, the fishes with the largest body sizes and ranges were predominately live-bearers, which have a limited dispersal capacity. This may explain why we found only partial support for predicted relationships in temperate marine fishes. If fishes are the exception to the rule then they are worth detailed examination to understand the mechanisms which ultimately drive these relationships. In addition, this research revealed that relationships between these variables were very different at upper and lower bounds, which may help to explain the diversity of relationship shapes that have been reported in the literature. Were we to only have explored these relationships by fitting regressions to the mean, constraint relationships would have been obscured. Future investigations of macroecological relationships should consider adopting a similar approach as it provides a more complete understanding of factors that shape species’ distributions and abundance.

LITERATURE CITED Ackerman JL, Bellwood DR (2003) The contribution of small individuals to density-body size relationships. Oecologia 136:137-140. Ackerman JL, Bellwood DR, Brown, JH (2004) The contribution of small individuals to density-body size relationships: examination of energetic equivalence in reef fishes. Oecologia 139:568-571. Bellwood DR, Meyer CP (2009) Searching for heat in a marine biodiversity hotspot. Journal of Biogeography 36(4):569-576. Blackburn TM, Gaston KJ (1996) Abundance-body size relationships: the area you census tells you more. Oikos 75:303-309 Blackburn TM, Gaston KJ (1997) A critical assessment of the form of the interspecific relationship between abundance and body size in animals. Journal of Animal Ecology 66:233-249. Blackburn TM, Lawton JH, Perry JN (1992) A method of estimating the slope of upper bounds of plots of body size and abundance in natural animal assemblages. Oikos 65:107- 112 Borregaard MK, Rahbek C (2010) Causality of the relationship between geographic distributio and species abundance. Quarterly Review of Biology 85:3-25 Bradbury IR, Laurel B, Snelgrove PVR, Bentzen P, Campana SE (2008) Global patterns in marine dispersal estimates: the influence of geography, taxonomic category and life

- 125 -

history. Proceedings of the Royal Society B-Biological Sciences 275:1803-1809 Brown JH (1984) On the relationship between abundance and distribution of species. The American Naturalist 124:255-279 Brown JH (1995) Macroecology, University of Chicago Press, Chicago Brown JH, Maurer BA (1987) Evolution of species assemblages - effects of energetic constraints and species dynamics on the diversification of the North American avifauna. American Naturalist 130:1-17 Cade BS, Noon BR, Flather CH (2005) Quantile regression reveals hidden bias and uncertainty in habitat models. Ecology 86:786-800 Carbone C, Rowcliffe JM, Cowlishaw G, Isaac NJB (2007) The scaling of abundance in consumers and their resources: implications for the energy equivalence rule. American Naturalist 170:479-484 Cofre HL, Böhning-Gaese K, Marquet PA (2007) Rarity in Chilean forest birds: which ecological and life-history traits matter? Diversity and Distributions 13:203-212 Colton MA, Swearer SE (2010) A comparison of two survey methods: differences between underwater visual census and baited remote underwater video. Marine Ecology-Progress Series 400:19-36 Compagno LJV (2001) Sharks of the world. An annotated and illustrated catalogue of shark species known to date. Volume 2: Bullhead, mackerel and carpet sharks (Heterodontiformes, Lamniformes and Orectolobiformes). In: FAO Species Catalogue for Fishery Purposes. Food and Agriculture Organization of the United Nations, Rome, p 269 Connell SD, Irving AD (2009) Integrating ecology with biogeography using landscape characteristics: a case study of subtidal habitat across continental Australia. Journal of Biogeography 35:1608-1621. Cotgreave P (1993) The relationship between body size and population abundance in animals. Trends in Ecology and Evolution 8:244-248 Damuth J (1981) Population density and body size in mammals. Nature 290:699-700 Edgar GJ (2005) Australian Marine Life: The Plants and Animals of Temperate Waters, Reed New Holland Publishers Pty Ltd, Sydney, Australia Eschmeyer WN, Fricke R (2009) Catalog of Fishes electronic version (9 September 2009). http://research.calacademy.org/ichthyology/catalog/fishcatmain.asp. Fisher JAD, Frank KT, Leggett WC (2010) Global variation in marine fish body size and its role in biodiversity–ecosystem functioning. Marine Ecology-Progress Series 405:1-13 Fowler AJ (1987) The development of sampling strategies for population studies of coral reef fishes - a case study. Coral Reefs 6:49-58 Francis M (2001) Coastal Fishes of New Zealand: An Identification Guide, Reed Books, Auckland, New Zealand Fretwell SD, Lucas Jr. HL (1969) On territorial behavior and other factors influencing habitat distribution in birds. Acta Biotheoretica 19:16-36 Gaston KJ, Blackburn TM, Lawton JH (1997) Interspecific abundance range size relationships: An appraisal of mechanisms. Journal of Animal Ecology 66:579-601 Gomon MF, Bray D, Kuiter R (2008) Fishes of Australia's Southern Coast, Reed New Holland, Sydney Hanski I (1991) Single-species metapopulation dynamics - concepts, models and observations. Biological Journal of the Linnean Society 42:17-38 Hanski I, Kouki J, Halkka A (1993) Three explanations of the positive relationship between distribution and abundance of species. In: Ricklefs RE, Schluter D (eds) Species Diversity in Ecological Communities: Historical and Geographical Perspectives. University of Chicago Press, Chicago, p 108-116 Harcourt AH, Coppeto SA, Parks SA (2005) The distribution-abundance (density) relationship: its form and causes in a tropical mammal order, Primates. Journal of Biogeography 32:565-579 Hawkins JP, Roberts CM, Clark V (2000) The threatened status of restricted-range coral reef fish species. Animal Conservation 3:81-88 Hijmans RJ, Guarino L, Jarvis A (2009) DIVA-GIS. Vsn. 7.0.5.1. A geographic information

- 126 -

system for the analysis of species distribution data. Manual available at http://www.diva- gis.org. Holling CS (1992) Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62:447-502 Hubble M (2003) The Ecological Significance of Body Size in Tropical Wrasses (Pisces: Labridae). PhD, James Cook University Johnson CN (1998) Species extinction and the relationship between distribution and abundance. Nature 394:272-274 Jones GP, Caley MJ, Munday PL (2002) Rarity in coral reef fish communities. In: Sale PF (ed) Coral Reef Fishes: Dynamics and Diversity in a Complex Ecosystem. Academic Press, San Diego, USA, p 81-101 Kattan GH (1992) Rarity and vulnerability - the birds of the Cordillera Central of Colombia. Conservation Biology 6:64-70 Kelt DA, Brown JH (1998) Diversification of body sizes: patterns and processes in the assembly of terrestrial mammal faunas. In: McKinney ML, Drake JA (eds) Biodiversity Dynamics: Turnover of Populations, Taxa, and Communities. Columbia University Press, New York, p 109-131 Kolb A, Barsch F, Diekmann M (2006) Determinants of local abundance and range size in forest vascular plants. Global Ecology & Biogeography 15:237-247 Kuiter RH (2000) Coastal Fishes of South-eastern Australia, Gary Allen Pty Ltd, Sydney Lancaster J, Belyea LR (2006) Defining the limits to local density: alternative views of abundance-environment relationships. Freshwater Biology 51:783-796 Lawton JH (1999) Are there general laws in ecology? Oikos 84:177-192 Lester SE, Ruttenberg BI (2005) The relationship between pelagic larval duration and range size in tropical reef fishes: a synthetic analysis. Proceedings of the Royal Society Biological Sciences Series B 272:585-591 McGill BJ (2008) Exploring predictions of abundance from body mass using hierarchical comparative approaches. American Naturalist 172:88-101 Mohr CO (1940) Comparative Populations of Game, Fur and Other Mammals. American Midland Naturalist 24:581-584 Mora C, Chitarro PM, Sale PF, Kritzer JP, Ludsin, SA (2003) Patterns and processes in reef fish diversity. Nature 421:933-936. Myers RA, Worm B (2003) Rapid worldwide depletion of predatory fish communities. Nature 423:280-283 Nee S, Gregory RD, May RM (1991a) Core and satellite species - theory and artifacts. Oikos 62:83-87 Nee S, Read AF, Greenwood JJD, Harvey PH (1991b) The relationship between abundance and body size in British birds. Nature 351:312-313 Neira FJ, Miskiewicz AG, Trnski T (1998) Larvae of Temperate Australian Fishes; Laboratory Guide for Larval Fish Identification, University of Western Australia Press, Nedlands, Western Australia Päivinen J, Grapputo A, Kaitala V, Komonen A, Kotiaho JS, Saarinen K, Wahlberg N (2005) Negative density-distribution relationship in butterflies. BMC Biology 3:13pp Pelc RA, Warner RR, Gaines SD (2009) Geographical patterns of genetic structure in marine species with contrasting life histories. Journal of Biogeography 36:1881-1890 Pérez-Matus A (2010) Effects of macroalgal habitats on the community and population structure of temperate reef fishes. PhD, Victoria University Peters RH, Raelson JV (1984) Relations between individual size and mammalian population density. American Naturalist 124:498-517 Pyron M (1999) Relationships between geographical range size, body size, local abundance, and habitat breadth in North American suckers and sunfishes. Journal of Biogeography 26:549-558 Quinn RM, Gaston KJ, Blackburn TM, Eversham BC (1997) Abundance-range size relationships of macrolepidoptera in Britain: the effects of taxonomy and life history variables. Ecological Entomology 22:453-461

- 127 -

Reaka ML (1980) Geographic range, life history patterns, and body size in a guild of coral- dwelling Mantis shrimps. Evolution 34:1019-1030 Root RB, Cappuccino N (1992) Patterns in population change and the organization of the insect community associated with goldenrod. Ecological Monographs 62:393-420 Rosenzweig ML, Lomolino MV (1997) Who gets the short bits of the broken stick? In: Kunin WE, Gaston KJ (eds) The Biology of Rarity. Chapman & Hall, London: p.63-90 Scott DW (1979) Optimal and data-based histograms. Biometrika 66:605-610 Smith WL, Wheeler WC (2006) Venom evolution widespread in fishes: A phylogenetic road map for the bioprospecting of piscine venoms. Journal of Heredity 97:206-217. Sturges HA (1926) The choice of a class interval. Journal of the American Statistical Association 21:65-66 Symonds MRE, Johnson CN (2006) Range size-abundance relationships in Australian passerines. Global Ecology and Biogeography 15:143-152 Tales E, Keith P, Oberdorff T (2004) Density-range size relationships in French riverine fishes. Oecologia 138:360-370 Taylor CC (1987) Akaike's information criterion and the histogram. Biometrika 74:636-639 Taylor CM, Gotelli NJ (1994) The macroecology of Cyprinella - correlates of phylogeny, body size, and geographical range. American Naturalist 144:549-569 Thomson JD, Weiblen G, Thomson BA, Alfaro S, Legendre P (1996) Untangling multiple factors in spatial distributions: Lilies, gophers, and rocks. Ecology 77:1698-1715 Warwick RM, Clarke KR (1996) Relationships between body-size, species abundance and diversity in marine benthic assemblages: facts or artefacts? Journal of Experimental Marine Biology and Ecology 202:63-71 Watson DL, Harvey ES, Anderson MJ, Kendrick GA (2005) A comparison of temperate reef fish assemblages recorded by three underwater stereo-video techniques. Marine Biology 148:415-425 Weersing K, Toonen RJ (2009) Population genetics, larval dispersal, and connectivity in marine systems. Marine Ecology-Progress Series 393:1-12 White EP, Morgan Ernest SK, Kerkhoff AJ, Enquist BJ (2007) Relationships between body size and abundance in ecology. Trends in Ecology and Evolution 22:323-330 Wiens JA (1989) Spatial scaling in ecology. Functional Ecology 3:385-397 Willis TJ (2001) Visual census methods underestimate density and diversity of cryptic reef fishes. Journal of Fish Biology 59:1408-1411 Willis TJ, Babcock RC (2000) A baited underwater video system for the determination of relative density of carnivorous reef fish. Marine and Freshwater Research 51:755-763

- 128 -

GENERAL CONCLUSION

It is well-recognized that the scale at which research is conducted will influence the patterns it reveals (Wiens 1989, Sale 1998). In this research, I investigated patterns in the distribution and abundance of temperate reef fishes at multiple spatial scales. At the smallest scale, I found that the method chosen to measure fish density will influence estimates of abundance and distribution (Chapter One). This underscores the necessity of selecting a sampling protocol that is appropriate to the study organisms. At a slightly larger scale, I revealed the importance of selecting a spatial grain appropriate to the size of an organism in order to investigate associations between fish and habitat (Chapter Two). At an even larger scale, however, the distribution of habitat became less informative, and currents, temperatures and habitat discontinuities offered a better explanation of species’ distributions (Chapter Three). At still larger scales, body size was found to place an ultimate constraint on abundance, and was found to be a good predictor of minimum geographic range size (Chapter Four). Somewhat surprisingly, geographic range size and abundance were negatively related in perciform fishes, and body size and geographic range size were negatively related in non-perciform fishes (Chapter Four). These data suggest that life history characteristics, in particular reproductive mode and dispersal capacity, may play an important role in shaping temperate marine fish communities at large scales.

Knowing where species occur and the sizes of their populations are essential data for management and conservation. This research provides estimates of these parameters for the nearshore rocky reef fish fauna of Victoria. As with all investigations, the results outlined in this dissertation suggest additional avenues of research. The biogeographic patterns revealed by this study highlight Victoria’s dynamic oceanography (Chapter Three), and suggest that phylogeographic studies of fishes in Victoria have the potential to be very informative. In particular, investigating the genetic structure of species with gaps in the centre of their ranges (Chapter Three) could address whether these disconnected populations are in fact the same species. Data on life history for most of the species included in this study were notably lacking. An understanding of pelagic larval durations coupled with an oceanographic model and information on thermal tolerances would allow for predictions to be made about how Victorian fishes might respond to climate change. Eastern Australia is

- 129 - predicted to be highly impacted by climate change (Figueira & Booth 2010), and in fact some range expansions associated with strengthening of the Eastern Australia Current have already been recorded in this region (Ling et al. 2009). While the estimates of abundance and distribution in this dissertation represent a snap-shot in time, it is hoped that the patterns revealed here will contribute much to our understanding of the factors structuring marine fish communities, allowing for better management and conservation in a changing environment.

LITERATURE CITED Figueira WF, Booth DJ (2010) Increasing ocean temperatures allow tropical fishes to survive overwinter in temperate waters. Global Change Biology 16:506-516 Ling SD, Johnson CR, Ridgway K, Hobday AJ, Haddon M (2009) Climate-driven range extension of a sea urchin: inferring future trends by analysis of recent population dynamics. Global Change Biology 15:719-731 Sale PF (1998) Appropriate spatial scales for studies of reef-fish ecology. Australian Journal of Ecology 23:202-208 Wiens JA (1989) Spatial scaling in ecology. Functional Ecology 3:385-397

- 130 -

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Colton, Madhavi A.

Title: Patterns in the distribution and abundance of reef fishes in South Eastern Australia

Date: 2011

Citation: Colton, M. A. (2011). Patterns in the distribution and abundance of reef fishes in South Eastern Australia. PhD thesis, Department of Zoology, The University of Melbourne.

Persistent Link: http://hdl.handle.net/11343/35914

File Description: Patterns in the distribution and abundance of reef fishes in South Eastern Australia

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.