UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evolution of the trophic niche and food web structure Ingram, Travis 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2011_fall_ingram_travis.pdf [ 3.68MB ]
JSON: 24-1.0105131.json
JSON-LD: 24-1.0105131-ld.json
RDF/XML (Pretty): 24-1.0105131-rdf.xml
RDF/JSON: 24-1.0105131-rdf.json
Turtle: 24-1.0105131-turtle.txt
N-Triples: 24-1.0105131-rdf-ntriples.txt
Original Record: 24-1.0105131-source.json
Full Text

Full Text

Evolution of the trophic niche and food web structure  by Travis Ingram B. Sc. Hons. Biology, University of Victoria, 2005  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  Doctor of Philosophy in THE FACULTY OF GRADUATE STUDIES (Zoology)  The University Of British Columbia (Vancouver) July 2011 c Travis Ingram, 2011  Abstract Food webs – networks of predator-prey interactions – are of fundamental importance to the ecological and evolutionary dynamics of biodiversity. The stability and functioning of food webs can be dependent on their ‘vertical’ structure: the distribution of species’ trophic positions, the length of food chains and the prevalence of omnivory. Food web interactions such as predation, resource competition and intraguild predation can be potent agents of natural selection, driving evolutionary responses that feed back to reconfigure the food web. The structure and function of food webs thus arises from an interplay of ecological and evolutionary processes. My thesis describes four studies of the evolutionary ecology of food webs. First, I test whether trophic position evolution is associated with speciation events in Sebastes rockfish. My phylogenetic comparative analyses find no signal of change at speciation in the evolution of trophic position or trophic morphology. Instead, speciation events in rockfish appear to be primarily associated with divergence in depth habitat in the ocean. Next, I use an evolutionary assembly model to explore how the strength of foraging trade-offs influences the structure and temporal dynamics of food webs, as well as patterns of trait evolution. Across a range of trade-off strengths, the amount of omnivory in a food web is positively related to both species turnover and the degree of convergence in trophic position evolution. I then fit macroevolutionary models to Sebastes trophic position data. The data support a model of recurrent evolution in a constrained trait space, as predicted for omnivorous consumers. Finally, I examine the ecological and evolutionary consequences of intraguild predation on threespine stickleback (Gasterosteus aculeatus) by prickly sculpin (Cottus asper). My collaborators and I use comparative and experimental studies to show that sculpin presence in lakes is associated with the evolution of antipredator and pelagic foraging morphology in stickleback, leading to reduced predator vulnerability, increased zooplanktivory, and changes to the structure of the food web. These studies address a number of important questions about how evolutionary processes influence food web structure and function, and illustrate the work that remains to be done in this exciting area of research.  ii  Preface Chapter 2 is modified from a publication in The Proceedings of the Royal Society of London: B, authored only by myself. Chapter 3 is modified from a publication in The American Naturalist, authored by myself, Luke J. Harmon and Jonathan B. Shurin. I led all stages of the research: developing the model, analyzing the results and writing the manuscript. LJH helped to write computer code, and LJH and JBS contributed to interpreting the results and writing the manuscript. Chapter 4 is authored only by myself. Chapter 5 is based on a manuscript authored by myself, Richard Svanb¨ack, Nathan J. B. Kraft, Pavel Kratina, Laura Southcott and Dolph Schluter. A portion of this chapter (comparative analysis of stickleback body shape) is based on data collected and analyzed by RS and DS. I led all stages of the larger, experimental part of the project: developing the hypotheses, conducting the research and analyzing the results in collaboration with NJBK, PK and DS. I also led the writing of the manuscript, which integrates the comparative and experimental data. RS, NJBK, PK, LS and DS contributed to writing and editing the manuscript. The experiment described in chapter 5 was approved by the University of British Columbia Animal Care Committee (permit # A07-0293). Publications related to thesis chapters: Ingram, T. 2011. Speciation along a depth gradient in a marine adaptive radiation. Proc R Soc Lond B 278:613–618. (Chapter 2) Ingram, T., L. J. Harmon and J. B. Shurin. 2009. Niche evolution, trophic structure and species turnover in model food webs. Am Nat 174:56–67. (Chapter 3)  iii  Table of Contents Abstract  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iii  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iv  List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vii  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  viii  Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ix  1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  1  1.1  Adaptive radiation and vertical structure of food webs . . . . . . . . . . . . . . . .  1  1.2  Intersections between food web ecology and evolution . . . . . . . . . . . . . . .  3  1.3  Study systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  4  1.3.1  Sebastes rockfish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  4  1.3.2  Stickleback and sculpin . . . . . . . . . . . . . . . . . . . . . . . . . . . .  5  Overview of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  5  1.4.1  Is trophic niche evolution involved in speciation? . . . . . . . . . . . . . .  5  1.4.2  How does trophic position evolve in the course of adaptive radiation? . . .  6  1.4.3  How does evolution modify food web interactions? . . . . . . . . . . . . .  6  1.4  2  Speciation along a depth gradient in a marine adaptive radiation  . . . . . . . . . .  7  2.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  7  2.2  Material and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  10  2.2.1  Phylogeny reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . .  10  2.2.2  Rockfish ecological and morphological data . . . . . . . . . . . . . . . . .  10  2.2.3  Inferring speciational vs. gradual evolution . . . . . . . . . . . . . . . . .  13  2.3  Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  14  2.4  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  16  iv  3  Niche evolution, trophic structure and species turnover in model food webs.  . . . .  20  3.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  20  3.2  Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  22  3.2.1  Model presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  22  3.2.2  Food web structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  25  3.2.3  Dynamics of species turnover . . . . . . . . . . . . . . . . . . . . . . . .  26  3.2.4  Evolutionary patterns of trophic position . . . . . . . . . . . . . . . . . .  27  Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  28  3.3.1  Food web structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  28  3.3.2  Dynamics of species turnover . . . . . . . . . . . . . . . . . . . . . . . .  30  3.3.3  Evolutionary patterns of trophic position . . . . . . . . . . . . . . . . . .  30  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  32  3.3  3.4 4  5  Tempo and mode of trophic position evolution in northeast Pacific rockfish (Sebastes) 37 4.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  37  4.2  Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  39  4.2.1  Stable isotope analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . .  39  4.2.2  Analysis of relationships between trophic position and morphology . . . .  42  4.2.3  Evolutionary analysis of trophic position . . . . . . . . . . . . . . . . . .  43  4.3  Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  45  4.4  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  49  4.4.1  Variation in trophic position in Sebastes . . . . . . . . . . . . . . . . . . .  49  4.4.2  Interpretation of evolutionary models . . . . . . . . . . . . . . . . . . . .  50  4.4.3  Potential drivers of trophic position evolution . . . . . . . . . . . . . . . .  51  4.4.4  Implications for the evolutionary recovery of food webs . . . . . . . . . .  53  Intraguild predation drives evolutionary niche shift by threespine stickleback  . . .  54  5.1  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  54  5.2  Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  56  5.2.1  Comparative analysis of body shape . . . . . . . . . . . . . . . . . . . . .  56  5.2.2  Mesocosm experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . .  59  5.2.3  Stickleback survival, growth and reproduction . . . . . . . . . . . . . . . .  60  5.2.4  Stickleback diet and impact on prey communities . . . . . . . . . . . . . .  61  5.2.5  Stickleback morphology . . . . . . . . . . . . . . . . . . . . . . . . . . .  62  Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  63  5.3.1  Comparative analysis of body shape . . . . . . . . . . . . . . . . . . . . .  63  5.3.2  Stickleback survival, growth and reproduction . . . . . . . . . . . . . . . .  65  5.3.3  Stickleback diet and impact on prey communities . . . . . . . . . . . . . .  65  5.3  v  5.3.4  Stickleback morphology . . . . . . . . . . . . . . . . . . . . . . . . . . .  68  Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  69  Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  73  6.1  Is trophic niche evolution involved in speciation? . . . . . . . . . . . . . . . . . .  73  6.2  How does trophic position evolve in the course of adaptive radiation? . . . . . . .  74  6.3  How does evolution modify food web interactions? . . . . . . . . . . . . . . . . .  75  6.4  Future directions in the study of trophic niche evolution . . . . . . . . . . . . . . .  76  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  78  A Appendix A: Sources of rockfish specimens . . . . . . . . . . . . . . . . . . . . . . .  92  5.4 6  vi  List of Tables Table 2.1  Associations between morphology, depth habitat and trophic position . . . . . .  16  Table 2.2  Phylogenetic analysis of speciational vs. gradual evolution . . . . . . . . . . .  17  Table 4.1  Regression model for predicting trophic position from morphology . . . . . . .  45  Table 4.2  Results of evolutionary model fitting of trophic position and morphological traits  48  Table 5.1  Lakes sampled for stickleback body shape analysis . . . . . . . . . . . . . . . .  57  vii  List of Figures Figure 2.1  Sebastes latitudinal and depth range overlap and phylogenetic distance . . . . .  9  Figure 2.2  Sebastes phylogeny showing latitudinal and depth ranges of each species . . .  11  Figure 2.3  Relationships between rockfish morphology, trophic position and depth habitat  15  Figure 2.4  Speciational evolution of depth but not trophic position . . . . . . . . . . . . .  17  Figure 3.1  Resource utilization curves and trade-off parameters . . . . . . . . . . . . . .  23  Figure 3.2  Trophic structure of model food webs . . . . . . . . . . . . . . . . . . . . . .  29  Figure 3.3  Dynamics of species richness and lineage accumulation . . . . . . . . . . . . .  31  Figure 3.4  Trophic structure, species turnover and evolutionary model fit . . . . . . . . .  32  Figure 3.5  Mapping of trophic position to phylogeny under different model scenarios. . .  33  Figure 4.1  Stable isotope data for rockfish and baseline organisms . . . . . . . . . . . . .  41  Figure 4.2  Estimated trophic positions and carbon sources of 32 Sebastes species . . . . .  46  Figure 4.3  Sebastes phylogeny showing species’ trophic positions . . . . . . . . . . . . .  47  Figure 4.4  Simulated waiting times until the evolution of a new top predator . . . . . . . .  50  Figure 5.1  Increased efficiency and niche width hypotheses . . . . . . . . . . . . . . . .  55  Figure 5.2  Landmarks used for morphometric analysis . . . . . . . . . . . . . . . . . . .  58  Figure 5.3  Shift in stickleback body shape associated with sculpin presence . . . . . . . .  64  Figure 5.4  Sculpin effects on stickleback survival and growth . . . . . . . . . . . . . . .  66  Figure 5.5  Differences in diet between stickleback from lakes with and without sculpin . .  67  Figure 5.6  Effects of niche shift on prey biomass in the food web . . . . . . . . . . . . .  67  Figure 5.7  Redundancy analysis showing effects of treatments on invertebrate composition  68  Figure 5.8  Morphology of sympatric and allopatric stickleback populations . . . . . . . .  70  viii  Acknowledgments I would first like to thank my two Ph.D. advisors. I thank Jonathan Shurin for taking a chance on me based on little more than a hunch; for his incessant enthusiasm; for subtly guiding me in the right direction when I needed it; and for his continued support from a distance. I thank Dolph Schluter for allowing me to join his lab late in my degree; for pushing me to think rigorously about the questions I ask; and for helping me to become a better teacher and scientist. Thanks to my supervisory committee, Wayne Maddison and Rick Taylor, for their advice over the years. I thank my colleagues past and present in the Shurin and Schluter labs and in the DeltaTea lab group for discussion and support during my time at UBC. I also would like to thank Tom Reimchen for piquing my interest in evolutionary ecology, and for the example he set as a natural historian and scientist. Many individuals helped me to successfully carry out the research described in this thesis. I thank John Hyde for sharing sequence data, Chad Brock for assistance with phylogeny reconstruction, and Rich FitzJohn for advice about trait evolutionary analysis. Thanks to Stefan Dick, Milton Love, Russ Markel, Katherine Maslenikov, Merit McCrea, Anne Salomon, Rick Taylor and Lynne Yamanaka for assistance with the compilation of my rockfish dataset. Thank you to Ashley Smith, Travis Tai, Anita Norman and Michaela Martin for field and lab assistance. Thanks to all those who I have worked with on my thesis research and on other projects during my time at UBC. Thanks for Dan Bolnick and his lab members for allowing me to join their field crew and for many intellectually stimulating discussions. I have benefitted tremendously from my collaborations with Arne Mooers, Mike Steel, Blake Matthews, Luke Harmon, Nathan Kraft, Pavel Kratina, B´eata Faller, Karen Magnuson-Ford and Becca Gooding. I thank my family for their support and love. Finally, I thank On Lee Lau, my colleague, field assistant, editor, and companion.  ix  Chapter 1  Introduction 1.1  Adaptive radiation and vertical structure of food webs  Adaptive radiation – the evolution of ecological and morphological diversity in a rapidly multiplying lineage – is responsible for much of global biodiversity (Simpson, 1953; Schluter, 2000). The ecological theory of adaptive radiation posits that this diversification is the result of divergent natural selection driving species to different niches, or peaks in the adaptive landscape (Schluter, 2000). Divergence often occurs along multiple ecological axes, but dietary niche differentiation is particularly important in many well-known radiations such as Darwin’s finches, African cichlids and threespine stickleback (Grant and Grant, 2006; Kocher, 2004; Schluter, 1994). Resource competition can promote diversification in diet, including niche expansion within a species or character displacement between species (Schluter, 2000). Dietary (trophic) niche evolution will impact not only the dynamics of the radiating lineage, but also its interactions with other species and thus the structure of the food web. Food webs are networks describing who eats whom in ecological communities. Direct links between species represent consumer-resource interactions, and combinations of links describe various indirect interactions such as resource competition and apparent (predator-mediated) competition. Natural food webs contain many species and many more direct and indirect trophic interactions, making them inherently complex biological systems. One way to ask ecologically meaningful questions about food webs in spite of this complexity is to focus on a single key dimension: the ‘vertical’ structure of the food web. The vertical dimension of food webs arises because of the directionality of energy and nutrient transfer from resources to consumers. This directionality allowed early ecologists to organize food webs into discrete trophic levels (Lindeman, 1942): energy and nutrients are taken in by producers at trophic level 1 and passed to herbivores (trophic level 2) and one or more carnivore trophic levels. The strict trophic level concept only applies when each species feeds  1  only on a single trophic level, as was suggested by early food web studies (Pimm, 1980). More detailed food web data later revealed that many species in fact are omnivores feeding at multiple trophic levels (Polis, 1991; Goldwasser and Roughgarden, 1993). The vertical structure of food webs can still be a useful organizing framework if we replace the strict trophic level concept with a continuous metric of trophic position, which measures the average number of trophic steps separating a species from the base of the food web (Levine, 1980; Vander Zanden and Rasmussen, 1996; Post, 2002b). Some species – particularly plants and herbivores – occupy discrete trophic levels (Williams and Martinez, 2004; Thompson et al., 2007), while many consumers have non-integer trophic positions between 2 and a maximum anywhere between 3 and 5.5, depending on the ecosystem (Post, 2002b; Vander Zanden and Fetzer, 2007). Many groups of organisms evolve little or no diversity in trophic position over long time periods. Most plant groups and many animal groups have been limited to horizontal diversification. Examples include seed-eating Geospiza finches, herbivorous parrotfishes, and many radiations of phytophagous insects that diversify between host plants. However, many other evolutionary radiations do involve – in some cases substantial – diversification in trophic position. The herbivorous Geospiza are closely related to other, insectivorous species of Darwin’s finches (Sato et al., 1999), while insectivorous Liolaemid lizards have recurrently evolved into herbivores (Espinoza et al., 2004). Many fish clades appear to evolve different trophic positions relatively rapidly (Jonsson, 2001; Matthews et al., 2010), the extreme case of which is the cichlid radiations in the African great lakes that include herbivorous, insectivorous, planktivorous and piscivorous species (R¨uber et al., 1999; Wagner et al., 2009). Trophic position evolution may have a number of causes. First, trophic position may evolve simply because it is correlated with traits under selection, such as body size. Alternatively, trophic position may be a more direct target of natural selection when there are fitness advantages associated with feeding on plants, herbivores or carnivores. One possible driver is the general increase in resource quality and decrease in quantity with trophic position (Fagan et al., 2002; Hendrixson et al., 2007). Organisms’ trophic positions may also change if they shift to habitats with longer or shorter food chains (Matthews et al., 2010), or as a result of the immigration, extinction or evolution of species lower in the food web (Post and Takimoto, 2007). To understand the evolution of trophic position we must also address the role of omnivory. Omnivory traditionally refers to a category of organisms that feed at multiple trophic levels (Pimm and Lawton, 1978), though it can be defined quantitatively as the variance of trophic positions of a consumer’s prey (Levine, 1980). The importance of omnivory comes from the low likelihood that a consumer will evolve from one trophic level to another without the intermediate step of feeding partially on each trophic level as an omnivore. Omnivory is widespread in food webs, although its importance varies among ecosystem types (Thompson et al., 2007). The presence of omnivory was once thought to destabilize food webs (Pimm and Lawton, 1978; 2  Pimm, 1980), but subsequent theory has shown that weakly interacting omnivores can be a stabilizing feature (McCann and Hastings, 1997; Vandermeer, 2006). The presence of omnivores can buffer food webs from strong predator-prey interactions (Bascompte et al., 2005), including trophic cascades (indirect positive effects of predators on basal resources resulting from depletion of their prey). As I will discuss later, a special class of omnivore preys upon a species with which it also competes for resources (i.e. the species share a trophic guild). This interaction – intraguild predation – is a promising system in which to investigate the rapid evolution of trophic position.  1.2  Intersections between food web ecology and evolution  Recent years have seen a growing appreciation for the existence of reciprocal interactions between evolutionary and ecological processes (Johnson and Stinchcombe, 2007; Post and Palkovacs, 2009; Schoener, 2011). There are a number of ways in which evolution may impact the structure and function of food webs. First, food webs are assembled of species whose characteristics are the result of past evolutionary processes: speciation, extinction and the evolution of traits that influence diet. Understanding what properties of organisms and environments contributed to the evolution of present-day trophic diversity may help us to develop strategies to preserve diverse food webs, including vulnerable high trophic position species. The growing appreciation that important evolution can occur on ‘ecological’ timescales raises the possibility that trophic interactions and evolutionary change can have reciprocal influences on one another (Hairston Jr et al., 2005; Carroll et al., 2007). It has long been appreciated that competition and predation are important agents of natural selection that can drive rapid evolutionary change (Reznick et al., 1990; Schluter, 1994; Langerhans et al., 2004; Grant and Grant, 2006; Nosil and Crespi, 2006). A number of recent experiments have demonstrated that evolutionary divergence driven by trophic interactions can impact the structure and function of food webs and ecosystems (Harmon et al., 2009; Bassar et al., 2010; Palkovacs and Post, 2009). A full understanding of food web dynamics may thus require knowledge of how evolutionary change and ecological interactions feed back on one another (Post and Palkovacs, 2009; Schoener, 2011). It is possible that evolution within populations that involves changes in trophic position could impact food web dynamics on short timescales. Non-evolutionary demographic changes can alter the trophic position of perch (Perca fluviatilis), and the presence or absence of large cannibalistic individuals changes the state of the lake food web via a trophic cascade (Persson et al., 2003). Evolutionary change may similarly alter the trophic position of omnivorous consumers, including cannibals and intraguild predators. The evolution of body size and other traits that influence trophic position has the potential to alter food web structure and ecosystem processes over short timescales.  3  1.3  Study systems  For my thesis work I use two empirical study systems: a diverse genus of marine fish (Sebastes rockfish), and a pair of interacting freshwater fish species (the threespine stickleback Gasterosteus aculeatus, and the prickly sculpin Cottus asper). Here I provide a brief introduction to the biology of these organisms; additional detail can be found in the relevant chapters.  1.3.1  Sebastes rockfish  There are more than 100 Sebastes species worldwide, and close to 70 in the northeast Pacific from Alaska to Baja California. The remaining species occur in the northwest Pacific (the area of the genus’ origination), the Gulf of California, the Atlantic Ocean and in the southern hemisphere. In the studies presented here, I focus on the 66 northeast Pacific species that were described at the time I began compiling my dataset. Almost all of these species appear to have originated in the northeast Pacific, so I treat this region as a roughly independent arena of diversification, and ignore the species from other regions. Sebastes is well-suited to testing hypotheses about trophic niche evolution due to its diversity in species and resource use (Love et al., 2002) and the existence of a well-resolved molecular phylogeny (Hyde and Vetter, 2007). Trophic niche evolution has not formally been studied in this group, although species are known to feed on different combinations of zooplankton, benthic invertebrates and fish (Love et al., 2002). Rockfish species span almost an order of magnitude of maximum body sizes, from the dwarf-red rockfish (S. rufinanus) at 18 cm total length to the shortraker rockfish (S. borealis) at 108 cm, providing ample scope for size-based variation in trophic position. While rockfish do not possess the extreme variation in trophic morphology seen in other groups of teleost fishes such as cichlids and labrids (Wainwright et al., 2004; Hulsey, 2006), species do vary considerably in the number and length of their gill rakers, an ecomorphological trait closely related to diet in many fishes. Rockfish are also notable for their diverse color patterns and for a number of their life history traits. Rockfish fertilization is internal, and follows elaborate courtship rituals that may involve visual, auditory and olfactory cues (Helvey, 1982). These characteristics may facilitate the evolution of assortative mating, potentially facilitating ecological speciation in the absence of strict geographic barriers. Following incubation, female rockfish undergo parturition, releasing thousands of recently hatched larvae into the water column (Love et al., 2002). After a pelagic phase of a few weeks to a few months, juvenile rockfish settle to shallow benthic habitats such as kelp forests and seagrass beds. As they mature, rockfish undergo an ontogenetic habitat shift to deeper waters, eventually settling at their characteristic adult depth (ranging from shallow subtidal habitats to 600 m). Rockfish mature very slowly, and can have extended lifespans up to 205 years (Love et al., 2002). These traits make some species highly vulnerable to  4  overexploitation, and many rockfish populations have declined in abundance and body size in recent decades (Parker et al., 2000; Love et al., 2002; Harvey et al., 2006). These threats highlight the importance of understanding the ecological and evolutionary factors that promote and maintain trophic diversity in Sebastes.  1.3.2  Stickleback and sculpin  My second empirical system consists of two fish species in freshwater food webs. Threespine stickleback are noteworthy for their extensive adaptive radiation in postglacial lakes and streams, including several instances of ecological speciation into sympatric benthic and limnetic specialist species pairs. These species pairs have been the subject of extensive research into the factors promoting speciation, including resource competition, predation, and ecological opportunity (Schluter and McPhail, 1992; Schluter, 1994; Rundle et al., 2003; Vamosi, 2003). Recent studies have shown that speciation in stickleback involves divergence in trophic position (Matthews et al., 2010) and can alter the structure and function of freshwater food webs (Harmon et al., 2009). In the work presented in this thesis, I focus on the trophic interactions and evolution of the ‘solitary’ stickleback populations (i.e. in lakes with a single stickleback species). In particular, I examine the interactions between stickleback and prickly sculpin, a benthic fish that both preys on stickleback (Moodie, 1972; Pressley, 1981) and competes with stickleback for benthic invertebrate prey (Bolnick et al., 2010), making it an intraguild predator of stickleback. Sculpin have the potential to exert strong selective pressures on stickleback through competition and/or predation, and the absence of sculpin appears to be a requirement for speciation in stickleback (Vamosi, 2003). As stickleback occur in multiple lakes with and without sculpin, this system is ideally suited to comparative and experimental tests of the effect of intraguild predators on intraguild prey evolution and food web structure.  1.4  Overview of thesis  In this document I present four studies that address three key questions about the causes and consequences of trophic position evolution.  1.4.1  Is trophic niche evolution involved in speciation?  In Chapter 2, I test whether divergence in trophic position is involved in speciation in Sebastes . I develop an evolutionary parameter that estimates the relative contribution of gradual and speciational change to a trait’s evolution. I contrast the evolutionary patterns of trophic position and trophic morphology with another important niche axis in rockfish: depth habitat. I find that depth habitat seems to consistently diverge at speciation throughout the radiation of Sebastes , but find no evidence for speciational evolution of trophic position or trophic morphology.  5  1.4.2  How does trophic position evolve in the course of adaptive radiation?  In Chapter 3 and Chapter 4, I attempt to link patterns of trophic position on phylogenies to the processes that underlie trophic position evolution. Chapter 3 is based on a simulation model of food web evolution, which I use to investigate how ecological parameters such as foraging trade-offs affect the structure of food webs, their dynamic stability, and the relationship between structure and stability. I also ask how the degree of omnivory in a clade is related to the macroevolutionary patterns of trophic position (i.e. how phylogenetic distance and trophic position divergence are related). The results of these allow me to develop predictions about how trophic position should map to a phylogeny. The results of these simulations indicate that a model of recurrent evolution within a constrained trait space should fit best for a highly omnivorous clade, while a model of decelerating evolution over time should fit for a clade with discrete trophic niches. In Chapter 4, I apply the insights from this model to test which of three competing evolutionary models best describes the evolution of trophic position in rockfish. I compare an unbounded random walk (Brownian motion), a model of early rapid evolution followed by stasis, and a random walk with some constraint that tends to return trait values to a central value. I find that the constrained model clearly provides the best fit to trophic position and trophic morphology in Sebastes, indicating recurrent trophic position evolution within a limited range of values. This is consistent with the predicted patterns for food webs with extensive omnivory. I use the evolutionary parameters derived from this model-fitting approach to model the potential evolution of a new top marine predator (i.e. trophic position ≥ 5) following their loss from the food web.  1.4.3  How does evolution modify food web interactions?  Finally, in Chapter 5, I move from this broad comparative approach to a more mechanistic study of trophic niche evolution using the intraguild predation interaction between stickleback and sculpin. I present comparative morphological evidence that sculpin presence is repeatedly associated with the evolution of more marine-like (armoured) and limnetic-like body shapes in stickleback populations. I then report a mesocosm experiment which evaluated the effects of sculpin on populations of stickleback that naturally occur in the presence or absence of sculpin. The experiment supported the comparative study by showing both reduced susceptibility to predation and greater pelagic foraging in the population sympatric with sculpin. This evolutionary change by stickleback appears to have altered the structure of the food web by impacting the zooplankton community and decreasing the trophic position of sculpin. This result, in concert with my other research, highlights the importance of understanding the interplay of ecological and evolutionary processes in structuring food webs.  6  Chapter 2  Speciation along a depth gradient in a marine adaptive radiation 2.1  Introduction  The scarcity of geographic barriers to gene flow in marine ecosystems poses a challenge to the traditional view of speciation, with its emphasis on geographic isolation (Palumbi, 1994). Fully allopatric speciation can occur across barriers such as the Isthmus of Panama (Knowlton et al., 1993), but these features are not common enough to explain all marine speciation events (Palumbi, 1994). A possible explanation is that ecological speciation – reproductive isolation resulting from divergent natural selection between ecological niches – is widespread in marine taxa (Rocha et al., 2005; Puebla, 2009). Reproductive isolation can arise as a byproduct of ecological divergence, and selection against the production of intermediate phenotypes can favour assortative mating and facilitate speciation. While ecological speciation can occur between strictly allopatric populations, divergent natural selection can also drive speciation in the face of gene flow (Schluter, 2001). Ecological speciation with gene flow may involve divergence in two main aspects of the niche, which have been distinguished in studies of trait-based community assembly (Pickett and Bazzaz, 1978; Ackerly and Cornwell, 2007). Species may diverge between macrohabitats or along environmental gradients (the β -niche), or partition local resources such as food and microhabitats (the α-niche). In the ‘habitat-first’ model, the initial divergence during speciation is between β -niches, with any α-niche divergence coming later (Diamond, 1986). One version of this idea is formalized in models of parapatric speciation along environmental gradients, caused by divergent natural selection and assortative mating (Doebeli and Dieckmann, 2003). Habitat-first speciation has been inferred from sister-species comparisons in groups including birds (Diamond, 1986; Richman and Price, 1992) and Lake Victoria cichlids (Seehausen et al., 7  2008). An alternative to habitat-first speciation occurs when disruptive selection favors divergence in the α-niche within a habitat. Theoretical models show that sympatric speciation can be driven by disruptive selection on resource-use traits combined with assortative mating based on ecological or marker traits (Dieckmann and Doebeli, 1999; Van Doorn et al., 2009). While compelling evidence for ‘within-habitat’ speciation is rare, it comes from α-niche divergence between young sympatric sister species (Barluenga et al., 2006). Sister species comparisons provide valuable information about modes of niche divergence during recent speciation, but neglect information about earlier speciation events. Molecular phylogenies contain a partial record of such events, allowing the predictions of habitat-first and within-habitat speciation to be tested at the scale of entire clades undergoing adaptive radiation. If certain characters diverge during speciation, the amount of evolutionary change in those traits should be proportional to the number of speciation events, not to the amount of time elapsed (Bokma, 2008). In a clade in which habitat-first speciation predominates, we can predict that traits associated with the β -niche will show this pattern of ‘speciational’ change. Conversely, if within-habitat speciation is common, traits related to the α-niche (e.g. trophic morphology) should exhibit speciational evolution. The marine rockfish genus Sebastes originated in the mid-Miocene (ca. 8 m.y.a.; Hyde and Vetter, 2007) and subsequently diversified to produce over 100 extant species. The distribution of related taxa suggest that Sebastes originated in the northwest Pacific (Hyde and Vetter, 2007), but the present centre of rockfish diversity is in the northeast Pacific. At least 66 species occur between Alaska and Baja California, up to 56 of which occur in broad sympatry (i.e. within 1 degree latitude) off southern California (Love et al., 2002). This considerable diversity in the absence of geographic barriers has sparked interest in the factors promoting speciation in Sebastes (Hyde and Vetter, 2007; Mangel et al., 2007; Hyde et al., 2008; Burford, 2009). A recent comprehensive molecular phylogenetic appraisal of the genus (Hyde and Vetter, 2007) indicates that most speciation occurs within oceanic regions. While a few long-distance dispersal events have allowed colonization of new regions, there is little evidence for allopatric speciation between ocean basins or sides of the Pacific Ocean (Hyde and Vetter, 2007). While rockfish have a pelagic larval phase and are capable of long-distance dispersal, local recruitment of larvae and site fidelity of adults can permit genetic structure within species (Johansson et al., 2008). Thus, isolation by physical distance might result in allopatric divergence between northern and southern populations (Burford, 2009). Contrary to the predictions of this hypothesis of allopatric speciation by latitude (Barraclough and Vogler, 2000), most close relatives show near-complete latitudinal range overlap (Figure 2.1). Ecological divergence over smaller scales may be more important for rockfish speciation than sheer physical distance. Species occupy characteristic depth habitats ranging from the intertidal to ∼600 m (Love et al., 2002), and the relatively low overlap of sister species depth 8  2  1.0  ●●  4  6  8  0.8 0.6  ● ●●  ●  ● ●●● ●● ●  0  Node Age (m.y.a.)  ● ●  0.0  0.2  ●  ●  0.4  Depth Range Overlap  ●  0  ●  0.2  1.0 0.8 0.6  ●  0.4  ●●● ●  0.0  Latitudinal Range Overlap  ●●● ●● ●● ● ● ●● ● ● ● ● ● ●  2  4  6  8  Node Age (m.y.a.)  Figure 2.1: Overlap in latitudinal range and depth range compared to time since divergence (node age) for pairs of Sebastes species (sister species denoted by filled symbols). Range overlap is calculated as the range shared by the two species divided by the smaller of the two ranges. In striking contrast to the predictions of allopatric speciation by divergence in latitude (Barraclough and Vogler, 2000), recently diverged species showed extensive overlap in latitudinal range, which was negatively related to node age (Mantel test, p < 0.001). In contrast, many recently diverged species showed low overlap in depth distribution, with no overall relationship between depth range overlap and node age (p = 0.23) distributions (Figure 2.1) is suggestive of parapatric speciation on a depth gradient (Hyde et al., 2008). Rockfish species also show extensive α-niche diversity, some feeding primarily on zooplankton and others consuming mostly benthic invertebrates or fish (Love et al., 2002). Coexisting species tend to have limited diet overlap (Hallacher and Roberts, 1985) and overdispersed foraging traits (Ingram and Shurin, 2009), suggesting a role for food competition in structuring rockfish assemblages. Within-habitat speciation is possible if these differences evolve in sympatry at the time of speciation. It has been suggested that Sebastes’ elaborate courtship rituals and internal fertilization make assortative mating likely (Helvey, 1982; Hyde and Vetter, 2007), potentially facilitating either habitat-first or within-habitat speciation in the presence of gene flow (Dieckmann and Doebeli, 1999; Doebeli and Dieckmann, 2003). I tested these alternative hypotheses by fitting evolutionary models with and without speciational change to rockfish trait data, using a species-level phylogeny of the 66 Sebastes species in the northeast Pacific (Figure 2.2).  9  2.2  Material and methods  2.2.1  Phylogeny reconstruction  I reanalyzed sequence data from a thorough investigation of rockfish phylogeny (Hyde and Vetter, 2007), using the program BEAST (Drummond and Rambaut, 2007) to produce a time-calibrated tree from 2 nuclear and 7 mitochondrial genes. I used a strong normal prior (3 ± 0.05 s.d. m.y.a.) on the divergence time between S. alutus and a clade of four Atlantic species (Hyde and Vetter, 2007) and a weakly informative lognormal prior (7 ± 0.5 log s.d. m.y.a.) on the root age of Sebastes based on the previous estimates and available fossil data (Hyde and Vetter, 2007). I ran 2 × 107 iterations, sampling every 2500 for a total of 8000 sampled trees, and discarded the first 25% as burn-in. I retained the maximum clade credibility tree (which maximizes the sum of the posterior probabilities of nodes) with mean branch lengths calculated from the posterior distribution. I also retained 100 trees evenly sampled from the posterior distribution to assess the robustness of parameter estimates to phylogenetic uncertainty. The full phylogeny included 99 species of Sebastes as well as four outgroup species (Hyde and Vetter, 2007). For this study I removed all but the 66 northeast Pacific Sebastes species. The tree also did not include three morphologically cryptic species that have been proposed since the initiation of this study (Hyde and Vetter, 2007; Burford, 2009; Hyde et al., 2008), but otherwise includes complete sampling of all northeast Pacific species. The resulting tree (Figure 2.2) was closely congruent with the original published tree based on the same sequence data (Hyde and Vetter, 2007).  2.2.2  Rockfish ecological and morphological data  I compiled published estimates of rockfish species maximum and minimum latitudinal ranges and common adult depth ranges (Love et al., 2002). These were used in calculations of latitudinal and depth overlap (Figure 2.1), while midpoint latitude and square-root transformed midpoint depth were used as characters in phylogenetic analyses. As a diet (α-niche) axis I estimated species mean adult trophic positions from stable nitrogen isotope ratios (δ 15 N) for 1-15 individuals from 44 species, sampled in the Santa Barbara Channel and in British Columbia. I calculated trophic position as TP = (δ 15 N- δ 15 Nbase )/3.4 + 2, where δ 15 Nbase is the average δ 15 N of a primary consumer (Mytilus californianus) and 3.4 is the average δ 15 N enrichment per trophic step (Post, 2002b). Estimates of trophic position that better account for marine isotopic baselines (see Chapter 4) did not qualitatively affect the results of this study. I measured morphology of 543 individual rockfish representing all 66 northeast Pacific species in the phylogeny. The majority of species were represented by 5-10 individuals (median 7), though intraspecific sample sizes ranged from 1-46. Sample sizes and sources of specimens  10  Latitude  Depth  GC  GC  GC GC  GC GC  X  X  SH SH  SH SH  GC GC  20  NA  NA  S macdonaldi S ruberrimus S melanostomus S aurora S phillipsi S gilli S diploproa S sinensis S melanosema S cortezi S peduncularis S semicinctus S saxicolaN S saxicolaS S atrovirens S carnatus S chrysomelas S maliger S caurinus S nebulosus S dallii S rastrelliger S auriculatus S elongatus S paucispinis S jordani S goodei S mystinus S entomelas S flavidus S serranoides S melanops S levis S babcocki S nigrocinctus S serriceps S rubrivinctus S ensifer S eos S chlorostictus S rosenblatti S helvomaculatus S simulator S rosaceus S umbrosus S lentiginosus S notius S oculatus S capensis S constellatus S exsul S spinorbis S ovalis S hopkinsi S moseri S rufinanus S rufus S pinniger S miniatus2 S miniatus1 S kiyomatsui S scythropus S trivittatus S vulpes S schlegelii S taczanowskii S pachycephalus S hubbsi S oblongus S joyneri S thompsoni S inermis S crameri S reedi S polyspinis S ciliatus S variabilis S baramenuke S mentella S norvegicus S fasciatus S viviparus S alutus S zacentrus S emphaeus S variegatus S wilsoni S proriger S matsubarae S flammeus S iracundus S borealis S brevispinis S aleutianus S melanostictus S glaucus S steindachneri S owstoni S minor  GC GC  X  X  NWP NWP NWP NWP NWP NWP NWP NWP NWP NWP NWP NWP  NWP NWP NWP NWP NWP NWP NWP NWP NWP NWP NWP NWP  NWP A A A A  NWP A A A A  NWP NWP NWP  NWP NWP NWP  NWP NWP NWP NWP  NWP NWP NWP NWP  45  70  0  300  600  Figure 2.2: Sebastes phylogeny used for the main analyses in this chapter. To the right of each species name is its latitudinal (degrees) and depth (m) ranges. Species not in the northeast Pacific dataset are labeled (NWP = Northwest Pacific; A = Atlantic; GC = Gulf of California; SH = Southern Hemisphere; X = recently described NEP species, no data) 11  for each species are given in Appendix A. I measured total length (TL), eye width, body depth, lower jaw length, pectoral fin length and longest gill raker length, and counted the gill rakers on the first gill arch. Lower jaw length and pectoral fin length did not show significant relationships with either niche axis, and were not analyzed here (but see Chapter 4). I averaged the left and right sides of fish for bilateral measurements to reduce error due to bending of the body, and performed blind repeated measurements of 10 arbitrarily chosen fish to ensure low measurement error (repeatabilities 0.98-0.999). I log-transformed all traits aside from gill raker number to linearize relationships among traits and to ensure normal distributions of raw trait values and residuals. To separate the effects of body size and size-independent morphology, one approach is to average all trait values within each species before performing a phylogenetic size adjustment. However, individuals within species varied extensively in size, intraspecific sample sizes were insufficient to calculate size-adjusted trait values within species, and the body sizes were not necessarily representative (due to potential biases in fish taken by fishers or museum collectors). As an alternative, for each trait I used the ‘contrast’ program in PHYLIP to calculate a phylogenetically corrected reduced major axis (RMA) slope against body size (OLS slope divided by correlation coefficient) that accounted for intraspecific variability and correlations (Felsenstein, 2008). I calculated the maximum likelihood intercept across all rockfish individuals given this RMA slope, calculated each individual’s trait value as its residual from this regression line, and calculated means and standard errors for these residuals within each species. These mean residuals were used as species trait values in subsequent analyses. Due to the potential biases in the sizes of measured fish, I used log-transformed maximum TL for each species (Love et al., 2002; Froese and Pauly, 2011) as a measure of body size. I tested for relationships between morphological traits and ecological variables to identify traits that may adapt species to either different prey types or different depth habitats. I calculated the maximum likelihood (ML) degree of phylogenetic signal (λsig ) in each morphological trait and niche axis (Freckleton et al., 2002; Harmon et al., 2008). I pruned the tree of all but the 44 species with trophic position estimates, then for each character I transformed the tree based on the estimated λsig and calculated phylogenetically independent contrasts (Felsenstein, 1985). I carried out a multiple regression of the contrasts in each morphological trait (including TL) against contrasts in depth habitat and TP (Table 2.1). I quantified the association of each trait with depth (relative to trophic position) as the partial R2 for depth as a predictor of the trait value, divided by the sum of the partial R2 ’s for both depth and trophic position. I predicted that traits would show a speciational pattern if the niche axis they are more strongly associated with is involved in speciation.  12  2.2.3  Inferring speciational vs. gradual evolution  I used a new parameter ψ to quantify the contribution of speciation to a trait’s total evolutionary rate. ψ has two advantages over a similar parameter κ (Pagel, 1997), which raises branch lengths to an exponent to scale between speciational (κ = 0; all branch lengths equal) and gradual evolution (κ = 1; branches proportional to time). First, ψ is derived from biologically interpretable parameters: the rates of gradual evolution and speciation, and the magnitude of evolutionary change at speciation. Second, a limitation of κ is that it neglects any nodes in the phylogeny that are hidden due to extinction. In the absence of a detailed fossil record we cannot know precisely where the hidden nodes are located in the tree, but it is possible to use estimated speciation (λ ) and extinction (µ) rates to infer where hidden nodes are most likely to occur, particularly on long branches toward the root of the tree (Bokma, 2008). I estimated λ and µ from the distribution of branching times in the phylogeny (Nee et al., 1994; Rabosky, 2006; Bokma, 2008). Maximum likelihood estimates were λ = 0.351 (95% confidence interval 0.271 0.447) and µ = 0 (0 - 0.103). If we assume λ and µ remain constant across the tree, we can estimate the number of hidden speciation events on a branch beginning at time t1 and ending at time t2 as Sh = λ a(t2 − t1 ). a is the probability that a lineage originating at time t left no extant descendents, averaged across the branch (Bokma, 2008). The estimate that µ = 0 (implying that there are no hidden speciation events) may be unreliable due to nonrandom extinction or incomplete sampling (Rabosky, 2010), but results were qualitatively unchanged when I assumed extinction rates up to 75% of the speciation rate. I used the (typically non-integer) expected Sh for estimation, though one could also sample integer Sh as part of a Bayesian estimation procedure (Bokma, 2008). I model evolution occurring both gradually (as a Brownian motion process) with rate parameter σa2 , and as step change at speciation, with change in the trait values of both daughter species drawn from a Gaussian distribution with variance σc2 . The rate of speciational evolution is thus λ σc2 , and the total rate of evolution is σt2 = σa2 + λ σc2  (2.1)  σa2 (t1 − t2 ) + σc2 (So + Sh )  (2.2)  The variance of change over a branch is  where So is the number of known speciation events affecting that branch, and Sh is the expected number of hidden speciation events. So is generally 1 for each branch, although if species present in the phylogeny are missing trait data (e.g. trophic position in this study), they can be pruned from the tree and deleted nodes accounted for by adding to the So of the affected branches. A  13  simple reparameterization of the above model (Equation 2.1) results in the new parameter ψ: ψ=  λ σc2 σt2  (2.3)  ψ, defined as the fraction of interspecific evolutionary divergence that is due to speciational change, ranges between 0-1, and can be compared among traits measured on different scales. To calculate the likelihoods of values of ψ and σt2 , I first calculate the phylogenetic variance-covariance matrix V for the phylogeny after transforming branch lengths following Equation 2.1. The diagonal of V represents the root-to-tip path length of each extant species, while off-diagonal elements represent the path length from the root to the most recent common ancestor of two species. Intraspecific variability or measurement error can be accounted for by adding the squared standard error of species mean trait values to the diagonal of V (Harmon et al., 2010). The ancestral state at the root of the tree is estimated as aˆ = (1 V−1 1)−1 (1 V−1 X), where X is the vector of species means and 1 is a column vector of ones (O’Meara et al., 2006). The likelihood function is the multivariate normal distribution of X, with expectation E(X) (a vector in which each value is a) ˆ and the variance-covariance matrix V. This function is used to calculate the likelihood L of the species trait data given ψ, σt2 , λ , µ, aˆ and the tree (Harmon et al., 2010; O’Meara et al., 2006). I identified the ML estimates of ψ and σt2 for each trait using the ‘subplex’ optimization function in R (R Development Core Team, 2009), and estimated approximate confidence intervals on ψ using profile likelihood. I compared evolutionary models using AICc (Akaike’s information criterion, corrected for sample size) and Akaike weights, which balance goodness of fit with model complexity (Anderson et al., 2000; Wagenmakers and Farrell, 2004). The model in which ψ takes its ML value has three parameters (a, ˆ σt2 and ψ), while the simpler Brownian motion model (ψ = 0) reduces to two (aˆ and σa2 ). To evaluate the sensitivity of this analysis to phylogenetic uncertainty, I estimated ψ for each trait on 100 trees sampled from the posterior distribution.  2.3  Results  Fish occurring in deeper habitats tended to have larger eyes – consistent with adaptation to low light availability (Warrant and Locket, 2004) – and smaller body depths. In contrast, species with higher trophic position tended to be larger, with fewer, shorter gill rakers, consistent with the consumption of larger prey (Table 2.1; Figure 2.3). Thus, although some traits showed some association with both niche axes, eye width and body depth are more related to the β -niche, while total length, gill raker number and gill raker length are more related to the α-niche. There was no tendency for α- or β -niche associated traits to show greater phylogenetic signal (Table 2.2). Phylogenetic analysis showed variation among characters in the relative importance of 14  ●  2  ●  ●  ●  0  ● ●  ● ● ●● ●  ●  ●●  ● ●● ● ● ● ● ● ●  ● ● ●  ●  ●  ●  ●  ● ●  ●  −2  Contrasts in Gill Raker Number  4  ●  ● ● ●  ● ●  ●  ●  ●  −4  ●  0.00  0.04 0.10  −0.04  ●  ● ●  0.0  ● ● ● ●  ●  ●  ● ● ●● ● ●  ●  −0.1  ●  ●  ● ● ●  ● ●  ●  ●  ● ●  ● ●  ● ●  −0.08  −0.04  0.00  0.05  ●  0.04  ● ● ● ●  ●● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ●  ●  ● ●  ●●  ●  −2  0  ● ●  −0.08  ●● ●  −0.04  0.10  ● ● ● ● ● ● ● ● ●  ●  ●  0.00  Contrasts in Body Depth  ●  ●  ●  ●●  6  ●●  ●● ● ● ● ●● ●● ● ● ● ●● ● ● ● ●● ● ●●● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ●●●● ● ● ●  ● ● ● ●  ● ● ●  ● ●  0.00  4  ● ●  ●●  ●  ●  2  −0.20  ●●  ● ●  ●  ●  ●  −0.10  0.08 0.04 0.00  ●  −0.04  Contrasts in Total Length  ●  ● ●●  ●● ● ●●  ● ●  ●  ● ●  ●  ●  ●  ●  −4  ● ●  ●  ● ● ● ●  ●  ●  ●  ●  ●  ●  0.00  ●  ●  ●  ●  ●  −0.05  ● ● ● ●  ●  Contrasts in Eye Width  ●  ● ●  −0.15  0.1  ● ●●  −0.2  Contrasts in Gill Raker Length  0.2  ●  −0.10  −0.08  0.04  ●  −4  Contrasts in Trophic Position  −2  0  2  4  6  Contrasts in Depth Habitat  Figure 2.3: Univariate relationships between phylogenetically independent contrasts of rockfish morphological traits and niche axes. Each trait is shown compared to the niche axis it is most associated with: trophic position for gill raker number, gill raker length and total length (left column), and depth habitat for eye width and body depth (right column).  15  Table 2.1: Associations between morphological traits and two niche axes: depth habitat and trophic position (TP) Morphological Trait Total Length Gill Raker Number Gill Raker Length Eye Width Body Depth  Niche Axis Depth TP Depth TP Depth TP Depth TP Depth TP  Slope -0.005 0.569** 1.59 -40.95*** 0.207** -2.433*** 0.083* -0.514 -0.078** 0.392  Standard Error 0.024 0.192 0.954 7.496 0.059 0.461 0.036 0.286 0.025 0.197  Partial R2 0.001 0.178 0.064 0.421 0.233 0.405 0.113 0.073 0.191 0.089  Depth Assoc. 0.01 0.13 0.37 0.61 0.68  *p < 0.05, **p < 0.01, ***p < 0.001 Depth Assoc. is the relative association with depth (vs. trophic position), calculated as partial R2depth / (partial R2depth + partial R2TP ) speciational and gradual evolution (Figure 2.4). There was a strong signal of speciational evolution in depth habitat and depth-associated traits, with approximately half of the total evolutionary rate for depth habitat and body depth estimated to occur at speciation. The model with speciational change was strongly preferred by AIC for depth habitat and body depth, and moderately preferred for eye width. Latidudinal distribution showed only a weak signal of speciational change, which was not supported over the simpler model of purely gradual evolution. Contrary to the predictions of within-habitat speciation, trophic position and trophic morphology showed no signal of speciational evolution (although confidence intervals were very wide for total length and trophic position). These results are robust to uncertainty about the phylogeny and the extinction rate (data not shown). The results for trophic position are also robust when recalculated using more accurate baseline isotope data (Chapter 4). There was effectively no support for speciational evolution for trophic position calculated for 32 species with isotope data (ψ = 0.033, wψ = 0.182), or for an expanded dataset of 66 species that also includes trophic positions predicted from morphology (ψ = 0.033, wψ = 0.215).  2.4  Discussion  The key result of this study is a strong signal of speciational evolution in the depth habitats of rockfish species, and in traits that appear to adapt species to different depths. This finding supports the hypothesis that rockfish speciate along a depth gradient (Hyde et al., 2008), and argues against several alternative models of speciation. Allopatric speciation between northern and southern populations has been inferred for a possible incipient species pair within the blue 16  1.0  ● ψ = 0.2  ●  Gill Raker Length  ●  Eye Width  Body Depth  Gill Raker Number  Total Length  Depth Habitat  Trophic Position  Latitude  0.4  0.6  ●  0.2  ψ (Relative Speciational Evolution)  0.8  ψ = 0.6  ●  ψ=0  ●  ● 0.0  ●  ●  ●  ●  ●  ●  ●  ●  0.00  ●  0.25  0.50  0.75  Relative Association with Depth  Figure 2.4: Estimated contribution of speciation to the total rate of evolution of each character (ψ). Filled circles show maximum likelihood estimates of ψ, and bars indicate 95% confidence intervals (± 1.92 log-likelihood units). Morphological traits are arranged on the x-axis by their association with depth habitat relative to TP (Table 2.1). The transformations of the Sebastes tree to the right show the branch lengths implied by three relevant values of ψ (0, 0.2 and 0.6).  Table 2.2: Phylogenetic analysis of speciational vs. gradual evolution Character Latitude Trophic Position Depth Habitat Total Length Gill Raker Number Gill Raker Length Eye Width Body Depth  Depth Assoc. – – –  λsig 0.72 0.28 0.79  σt2 8.89 0.029 1.80  ψ 0.093 0 0.644  AICψ 422.3 40.3 350.3  AICψ=0 421.2 38.1 358.4  wψ 0.36 0.25 0.98  0.01 0.13 0.37 0.61 0.68  0.94 0.87 0.78 0.99 0.62  0.045 4.63 0.022 0.0024 0.0013  0 0 0 0.226 0.498  63.3 372.1 26 -106.4 -123.7  61.1 369.9 23.8 -105.6 -111.2  0.25 0.25 0.25 0.61 1.00  Depth Assoc. is the relative association with depth (vs. trophic position; Table 2.2). Maximum likelihood estimates (MLE) are given for λsig (strength of phylogenetic signal), σt2 (total evolutionary rate) and ψ (proportion of speciational evolution). AIC values are given for models with the MLE of ψ and for Brownian Motion (ψ = 0), and the Akaike Weight wψ gives the relative support for the model with the ψ term included. 17  rockfish, S. mystinus (Burford, 2009), but the weak signal of speciational change in latitudinal distribution suggest that fully allopatric speciation is rare in northeast Pacific Sebastes. It has also been proposed that rockfish speciation involves divergence in life history, particularly in maximum lifespan, which varies from 10-200 years among species (Mangel et al., 2007). The lack of a speciational signal in total length (a strong correlate of lifespan) suggests that this is not a general feature of rockfish speciation. Finally, an alternative explanation for a signal of speciational change in traits is that features of the speciation process (e.g. population bottlenecks) lead to abrupt change in many characters (Bokma, 2008). The fact that traits unrelated to depth do not show a signal of speciational change is inconsistent with such a punctuated equilibrium model, and indicates a special role for depth habitat in the speciation process. These findings also draw intriguing connections between the processes of trait evolution, speciation and community assembly. The concepts of the α- and β -niche originated in studies of community assembly, and were developed to describe how species sort themselves within and between habitats, respectively (Pickett and Bazzaz, 1978; Ackerly and Cornwell, 2007). When the assemblages of interest are also members of a clade undergoing adaptive radiation, we can extend this framework to consider how species diversity accumulates via speciation, with the distinct modes of speciation corresponding to the α- and β -niche as described in this manuscript. In the case of Sebastes, the β -niche appears to be involved in speciation, while the α-niche does not. Some researchers have demonstrated a burst of α-niche evolution early in the diversification of a clade, followed by α-niche conservatism (Ackerly et al., 2006; Richman and Price, 1992; Glor et al., 2003). In contrast, I found no indication that the α-niche exhibits greater conservatism (i.e. phylogenetic signal) than the β -niche: trophic position had relatively low phylogenetic signal, while morphological traits more associated with depth vs. trophic position did not differ consistently in phylogenetic signal. This pattern implies that both α- and β -niche evolution have occurred throughout the radiation, but that only β -niche evolution has been concentrated at speciation. Although divergence in trophic position may not be involved in speciation, α-niche diversification may nonetheless contribute to the species diversity of Sebastes. If species initially isolated by depth habitat later evolve α-niche differences (due to any combination of adaptation to new resources, competition and genetic drift), they may compete less upon secondary contact and may subsequently be able to coexist (Ingram and Shurin, 2009). The speciational signal in β -niche evolution reported here suggests that habitat divergence can be an important component of speciation in marine taxa that have little opportunity for strict geographic isolation (Palumbi, 1994). Divergence in depth may result from multiple, not mutually exclusive processes. Rockfish recruit from the plankton to shallow habitats as juveniles, and later undergo ontogenetic migrations to their adult depth habitat. Individuals may recruit by chance to different depth habitats, especially when features such as offshore seamounts lead to a 18  discontinuous depth gradient (Hyde et al., 2008). Individuals may also recruit to new depth habitats created by sea level fluctuations (Hyde and Vetter, 2007). This spatial separation may result in individuals reproducing, and populations establishing, in separate habitats. Alternatively, intraspecific variation in traits that affect fitness in different depth habitats may result in divergent selection between shallower and deeper parts of a species depth range. Either selection or behavioral matching of phenotype to habitat may result in a correlation between depth and morphology within species. Migration of suboptimal phenotypes or competition between phenotypes may favour assortative mating, reducing gene flow and ultimately allowing speciation along the depth gradient (Doebeli and Dieckmann, 2003). However speciation by depth habitat occurs, it will result in sister species experiencing different light environments. This presents a further similarity between Sebastes and the cichlid radiations of Africas rift lakes (Johns and Avise, 1998). Like cichlids, rockfish appear to diverge in light environment and perhaps colour (Love et al., 2002) during speciation, and to undergo adaptive evolution of visual pigment genes along with habitat shifts (Sivasundar and Palumbi, 2010). In cichlids, detailed studies have demonstrated that speciation involves sensory drive, the integrated divergence of sexual signals and preferences between environments (Seehausen et al., 2008). While much work remains to identify the detailed mechanisms involved in rockfish speciation, these similarities suggest that Sebastes is a promising system for future studies of the role of both ecological speciation and sensory drive in marine adaptive radiations.  19  Chapter 3  Niche evolution, trophic structure and species turnover in model food webs. 3.1  Introduction  A considerable challenge facing community ecologists is to understand the causes of variation in trophic structure among ecosystems and the consequences for food web dynamics. Species foraging behavior and dietary niche breadths strongly influence food web structure (Kondoh, 2003; Beckerman et al., 2006). Food webs can consist of a few discrete trophic levels when most consumers specialize on a small number of trophically similar prey species, or the presence of many omnivores can lead to food webs with complex trophic structure. Recent meta-analyses have shown that virtually all food webs include both omnivores and species at discrete trophic levels, but that the relative prevalence of omnivory and trophic levels varies across ecosystem types (Williams and Martinez, 2004; Thompson et al., 2007). The trophic structure of food webs may influence their stability in a number of ways. Omnivory-stability relationships have been the subject of considerable research and occasional controversy (Vandermeer, 2006). Early theory suggested that omnivory decreases the probability of stable equilibria and therefore should be rare in nature (Pimm and Lawton, 1978; Pimm, 1982). This prediction appeared to be supported by early empirical data (Pimm, 1980), but later studies with improved resolution showed many species feeding on multiple trophic levels (Polis, 1991; Goldwasser and Roughgarden, 1993). More recent analytical models have identified conditions under which omnivory stabilizes food webs both by conferring a stable equilibrium and by reducing the probability of extinctions (McCann and Hastings, 1997; Vandermeer, 2006). Most current models consider only small modules of a few interacting species, making their implications for complex natural food webs unclear. Omnivory may stabilize large food webs by reducing the likelihood that strong predator-prey interactions will drive population cycles or 20  trophic cascades (strong top-down effects across more than two trophic levels). Fagan (1997) provided experimental support for this hypothesis by showing that arthropod assemblages with omnivores were more resilient to disturbance. Bascompte et al. (2005) indirectly demonstrated a stabilizing role of omnivory by showing that food chains with consecutive strong interactions (where trophic cascades are likely) were underrepresented in a large marine food web. When they occurred, consecutive strong interactions tended to include an omnivorous link between the top and basal species, suggesting that omnivory acts as a buffer against strong top-down control. While omnivory may have direct effects on food web stability, we suggest that omnivory and aspects of stability may be related if they are influenced by the same environmental features. Theory predicts that omnivory is most likely to be stabilizing if trophic interactions are weak, consistent with other links between weak interactions and food web stability (McCann et al., 1998; Emmerson and Yearsley, 2004). However, while a preponderance of weak interactions may favor the persistence of a food web, it may also increase community invasibility (Case, 1990). Omnivory may therefore have a complex relationship with the overall stability of a food web, favoring the maintenance of a stable trophic structure but facilitating turnover through species invasions and extinctions. To our knowledge, no studies have yet considered the relationship between levels of omnivory and components of food web stability such as invasibility and species turnover. We explore patterns of trophic structure and stability that arise through speciation-extinction dynamics using an evolutionary assembly model modified from Loeuille and Loreau (2005) (L&L). Evolutionary assembly models allow food webs to emerge from the evolution of traits that affect fitness via species interactions (Caldarelli et al., 1998; Yoshida, 2003; McKane, 2004; Rossberg et al., 2005; Guill and Drossel, 2008), in contrast to stochastic descriptive models of food webs (Cohen and Newman, 1985; Williams and Martinez, 2000; Cattin et al., 2004). In an attempt to directly link the assembly process to a measurable characteristic of species, L&L devised a model based on the evolution of a single measurable trait, body size. In L&L’s model, body size determines each species metabolic rates as well as the identity of its predators, prey and competitors. L&L’s model produced either complex networks of omnivores or food webs with all species at distinct trophic levels, depending largely on whether species can eat a wide or narrow range of prey (i.e. their dietary niche width). L&L’s model assumes that all species have identical niche widths, which may limit the model’s predictive ability given the marked variability in dietary generality in real food webs. The evolution of niche width is likely to have implications for both the fitness of species and for the structure and stability of the food web as a whole. Here, we extend L&L’s model by allowing niche width to evolve under a variety of trade-off scenarios. We assess the conditions that favor different types of trophic structure and consider how the assembly process may lead to emergent relationships between omnivory and the dynamic stability of food webs. We also use evolutionary models to examine the distribution of 21  trophic positions on a phylogenetic tree, allowing us to make predictions about how comparative data may relate to ecological factors associated with food web assembly.  3.2 3.2.1  Materials and methods Model presentation  We modify the model presented by Loeuille and Loreau (2005) by allowing the evolution of both body size and niche width. Each species i has a biomass density Ni and is characterized by two traits: body size (xi , on a logarithmic axis) and niche width (si ; see below). The basal resource N0 represents the inorganic nutrient pool, and is arbitrarily assigned a body size of x0 = 0. Predation is size-structured so that predators are always larger than their prey. Each consumer i has an optimum prey size of xi − d, with d = 2 following L&L and consistent with typical empirical predator-prey size ratios (Brose et al., 2005, 2006b). Consumers have Gaussian utilization functions with standard deviation (niche width) si . The attack rate ai j of consumer i on prey j (where xi > x j ) is: ai j = a0  exp −c(log(si /s0 ))2 √ si 2π  exp −  (xi − x j − d)2 s2i  (3.1)  Here, a0 scales the maximum attack rate, while c and s0 describe a trade-off associated with niche width. s0 is the optimum niche width at which overall attack rate is highest (s0 > 0), while c is the cost associated with deviating from this optimum (c ≥ 0). When c = 0, s can vary with no intrinsic fitness cost, as the numerator of the first fractional term in Equation 3.1 simplifies to 1 and total attack rate (the integral of Equation 3.1 over all x j from 0 to xi ) is independent of niche width (with the minor exception that species with large niche widths have slightly lower total attack rates as they are not permitted to consume species larger than themselves). This produces a commonly employed trade-off where increasing efficiency at exploiting a particular prey size decreases the range of body sizes that can be consumed (Roughgarden, 1972; Slatkin, 1980; Taper and Case, 1985) (Figure 3.1A). Niche width is unlikely to evolve unconstrained, and a review of experimental data shows that costs can be associated with both generalization and specialization (Kassen, 2002). We thus modify the utilization function so that a cost can be imposed to having a non-optimal niche width (for other approaches to constraining niche width evolution see Yoshida, 2003 and Ackermann and Doebeli, 2004). We assume a cost to having either a higher or lower niche width than the optimum s0 , allowing us to explore the effect of varying the optimum to give specialists or generalists an intrinsic advantage. As the cost c increases, the total attack rate decreases more rapidly when s differs from s0 (Figure 3.1B); conceptually, this scales down the utilization curve in Figure 3.1A so that the area under the  22  Scaled Total Attack Rate  d d  j  k  0  1  2  3  4  0.0 0.2 0.4 0.6 0.8 1.0 1.2  0.0 0.2 0.4 0.6 0.8 1.0 1.2  Attack Rate  A  B  *  *  0.5  1.0  i  5  6  0.0  Body Size (x)  1.5  2.0  2.5  Niche Width (s)  Figure 3.1: (A) Illustration of utilization curves for species differing in body size (x) and niche width (s). Species j (x j = 4, s j = 0.5, solid lines) consumes prey species k (xk = 2.0) with high efficiency, but species i (xi = 4.9, si = 1.0, dashed lines) can consume both j and k with lower efficiency. (B) Effect of a species niche width and the parameters c and s0 on total attack rate (i.e. the area under the curve in A, scaled to the maximum value. Curves show the relationships between total attack rate and niche width at three values of the cost parameter c (solid lines = 0, dashed lines = 0.5, dotted lines = 4) and two values of the optimum niche width s0 (0.5 and 1, indicated with asterisks). curve decreases as |s − s0 | increases. We infer reasonable narrow and wide optima (s0 = 0.5 and 1.0, respectively) from empirical consumer-resource body size data (Brose et al., 2006a; Barnes et al., 2008; Ingram et al., 2009). We chose values of c to capture its range of behavior (c = 0, 0.5 and 4). We calculate the total amount of each prey j consumed by i using a flexible functional predator response to prey density that applies to multispecies communities (modified from Drossel et al., 2004): gi j =  ai j N j 1 + ∑k aik Nk b  (3.2)  We use a weakly saturating Type II functional response (b = 0.1) for the simulations presented here, but explore the effects of Type I (b = 0) and strongly saturating Type II (b = 0.5) functional responses on our conclusions. Like previous evolutionary assembly models, the present model requires predator interference to sustain food webs of more than a few species (Drossel et al., 2004; Loeuille and Loreau, 2005; Guill and Drossel, 2008). Interference competition promotes diversity in this  23  model by intensifying interactions within species and between species with similar trophic positions. Interference is often incorporated into predator-dependent functional responses, but we follow L&L by making the simplifying assumption that interference occurs on the basis of body size. Interference may therefore occur either during foraging (as similar-sized species will have common prey) or in competition for other resources such as territories (Oksanen et al., 1979; Bowers and Brown, 1982). We use a Gaussian competition function so that competition is strongest between equally sized species and declines as the size difference increases: αi j =  (xi − x j )2 α0 √ exp − σα2 σα 2π  (3.3)  α0 determines the overall strength of interference competition and σα determines the width of the interference function: both values were assigned small values so that interference is weak relative to predation and only occurs between species with fairly similar body sizes. L&L thoroughly explored how the strength of interference competition influences food web structure in this framework, so we do not present the effects of varying α0 and σα here. Consistent with the results of L&L, we found that when α0 = 0 food webs contain few species and frequently collapse, while higher α0 tends to disrupt trophic levels and lead to a more uniform distribution of body sizes. We calculate population dynamics using a system of discrete-time recursion equations: Ni (t + ∆t) = Ni (t) + Ni (t) · growth · ∆t  (3.4)  For each species i aside from the basal resource, growth rate depends on its total prey assimilation and losses due to mortality, predation, interference competition: i−1  growth = f0 xi−0.25 ∑ N j gi j − m0 xi−0.25 − j=0  n  ∑  n  j=i+1  N j g ji − ∑ N j αi j  (3.5)  j=1  f0 and m0 are basal metabolic parameters (mass-specific production efficiency and mortality rates) that are scaled as an exponent of body size (Kleiber, 1947; Peters, 1983), and the other terms are as described above. The dynamics of the basal inorganic resource N0 are: n  Ni (t + ∆t) = N0 (t) I − e N0 + ν Nrecycled − ∑ Ni N0 gi0 ,  (3.6)  i=1  where n  n  n  Nrecycled = ∑ m0 xi−0.25 Ni + ∑ ∑ Ni N j αi j + (1 − f0 xi−0.25 )Ni N j gi j , i=1  i=1 j=1  I is the nutrient inflow at each time step, e is the rate of nutrient outflow and v is the rate of 24  (3.7)  nutrient recycling from higher trophic levels (L&L). We use a step size ∆t = 0.2 that balances low numerical instability with quick computation time (Drossel et al., 2001). New species are introduced to the system by stochastic speciation events with concurrent mutations in trait values. Speciation occurs with probability 0.005 in each generation, with one extant species randomly selected as the parent. The body size and niche width of the new species are randomly drawn from normal distributions around the parent species traits, with standard deviations σx and σs and no correlations between mutations in x and s. We present simulations in which mutations in both traits have either larger (σx = 0.5 and σs = 0.25) or smaller (σx = 0.2 and σs = 0.1) standard deviations. If undefined values of either trait are selected (x or s ≤ 0) both trait values are rejected and new values are drawn from the same distribution. Species are introduced at a low density (Ni = xi · 10−8 ), also the threshold density below which species are considered extinct and removed from the community. If new species have non-negative population growth (growth ≥ 0 in Equation 3.5) they successfully establish in the community; otherwise they immediately go extinct. We initialize each simulation with a single species that consumes the basal resource, with its x and s randomly drawn from normal distributions with expectations d and s0 and standard deviations σx and σs , respectively. Each simulation lasts 106 time steps, typically long enough for the food web to reach a dynamic equilibrium state where the number and type of species present undergoes little further directional change (Figure 3.3). We record species richness every 1000 generations and track ancestor-descendent relationships among species. For the results presented here, we varied three parameters, simulating food webs with: small and large optimum niche widths (s0 = 0.5, 1.0); no costs, weak costs and strong costs to deviating from the optimum (c = 0, 0.5, 4); and large and small trait mutations (σx = 0.5 and σs = 0.25, σx = 0.2 and σs = 0.1). We replicated each of these 12 parameter combinations 20 times to assess the consistency of the resulting food web structures. Simulations were carried out using code written in C, and subsequent data manipulations were performed in the R environment (R Development Core Team, 2009). We began by characterizing the trophic structure of food webs produced by our model. We then assessed food webs dynamical stability using a measure of variability and species turnover and a component of phylogenetic tree shape. Finally, we evaluated whether trophic structure, stability, and patterns of trait evolution are related across a range of simulated food webs.  3.2.2  Food web structure  Our analysis of food web structure focused on the distribution of trophic positions and levels of omnivory. We calculated trophic position (TP) and a measure of omnivory for each species following Levine (1980). A species’ TPi is the average number of trophic steps that separate it from the base of the food web (the basal resource, with TP = 0), while the trophic height of a 25  food web is the maximum TP of any species (TPmax ). Each species’ omnivory is the weighted variance of the trophic positions of its prey: n 2 σTP = i  ∑ (TP j − TPi − 1)2 pi j ,  (3.8)  j=1  where TP j are the trophic positions of species i’s prey and pi j is the proportion of each species j in the diet of species i (Levine 1980). We calculated the level of omnivory in a food 2 of all species present. web as the mean σTP  3.2.3  Dynamics of species turnover  We used a variety of methods to investigate assembly dynamics in our simulations. Evolutionary assembly models simulate adaptive radiations as species evolve to occupy new trophic niches. By tracking speciation and extinction over time we can construct a community phylogeny and investigate relationships between phylogenetic tree shape and ecological factors affecting species interactions. We examined the effects of varying the niche width trade-off (c and s0 ) and the rates of trait evolution (σx and σs ) on aspects of food web stability and phylogenetic tree shape. Food web stability has been defined in many ways, including analytical stability (tendency for all species to return to a stable equilibrium after slight perturbations; e.g. May, 1973) and robustness (susceptibility to cascading extinctions; Dunne et al., 2002; Ingram and Steel, 2010). One component of stability that is of particular relevance to evolving food webs is the degree of species turnover: the rates at which species invade via speciation and go extinct. We calculated the proportion of mutations that produced new species with initially positive growth rates (i.e. species that persisted in the food web for at least one generation) as a measure of species turnover. Alternative measures of temporal stability, such as the magnitude of fluctuations in species richness over time, led to similar conclusions. We examined patterns in phylogenetic tree shape using the phylogenies of species alive at the end of each simulation (after pruning out extinct lineages). Temporal patterns of branching events in phylogenies are typically presented as lineage-through-time (LTT) plots, with the reconstructed number of lineages (ignoring extinct species) against time. The shape of these LTT plots can be conveniently described using the gamma statistic (Pybus and Harvey, 2000), which is negative when most lineages diverge early in the phylogeny and positive when most branching events are recent. This statistic is typically used to test for changes in speciation rate over time (Phillimore and Price, 2008; Rabosky and Lovette, 2008), but has also been used in metacommunity simulations to test for effects of ecological parameters on tree shape (McPeek, 2008). We use gamma to ask whether branching events separating extant species tend to be early or more recent. We note that our simulations do not satisfy the conditions under which gamma is standard-normally distributed (Pybus and Harvey, 2000), but that the statistic remains a useful 26  descriptor of the shape of LTT plots. We calculated gamma for each phylogeny using the function ‘gammaStat’ in the R package ‘ape’ (Paradis et al., 2004). We investigated whether the trophic structure of the resulting food web (mean omnivory) was related to species turnover across simulations under a range of parameter values. While this analysis does not demonstrate causality (e.g. the presence of omnivory increasing or decreasing stability), it is a useful exploration of relationships that may emerge from repeating the assembly process across a variety of ecological conditions.  3.2.4  Evolutionary patterns of trophic position  We then used evolutionary model fitting to examine how species’ trophic positions were distributed across the phylogeny at the end of the simulation. Our motivation for this analysis was to generate hypotheses to assess how properties of a clade – such as degree of omnivory – predict the evolutionary dynamics of trophic position. These predictions are used in a subsequent chapter (Chapter 4) and may prove useful for future evolutionary studies of trophic position. We consider three simple evolutionary models that may describe patterns of trophic position diversity in clades; these are discussed in more detail in Chapter 4. The simplest model is Brownian motion (BM), in which traits evolve following a random walk consistent with genetic drift or to adaptation to a randomly fluctuating optimum (Felsenstein, 1985). Under this model, the variance among independently evolving lineages increases linearly with time, contrary to some empirical studies (Ackerly, 2009; Harmon et al., 2010). We therefore evaluate two alternative models that may apply to traits in which disparity does not continue to increase indefinitely over time (as is reasonable to expect for trophic position). A rapid early filling of trait space may be followed by stasis, as in classic models of adaptive radiation (Simpson, 1953). We model this scenario using the ‘Early Burst’ (EB) model, in which trait evolution follows a random walk whose rate declines exponentially over time (Blomberg et al., 2003; Harmon et al., 2010). Alternatively, the rate of trait evolution may remain relatively high, but one or more forces may constrain the trait space that is explored. The Ornstein-Uhlenbeck (OU) model combines a random walk with a deterministic tendency for trait values to return to an ‘optimum’ intermediate value, and can describe stabilizing selection around a single stationary peak or adaptation to a moving optimum whose position is constrained (Hansen and Martins, 1996; Hansen, 1997; Estes and Arnold, 2007; Harmon et al., 2010). We used maximum likelihood estimation to fit the BM, EB and OU models using the ‘fitContinuous’ function in the R package ‘geiger’ (Harmon et al., 2008). The units of the branch lengths (generations) posed difficulties for parameter estimation, so we arbitrarily multiplied all branch lengths by 10−5 . We excluded simulations that had effectively no variation among species in trophic position (TPmax < 1.1; this occasionally occurred in simulations with c = 0 and low σx and σs ), leaving 222 simulations with between 7 and 69 taxa each. We compared model fit using 27  small sample size-corrected Akaike’s Information Criterion (AICc ) (Anderson et al., 2000; Wagenmakers and Farrell, 2004) to account for differences in model complexity (2 parameters for the BM model, 3 for the EB and OU models), and compared the support for different models with the degree of omnivory in each simulation.  3.3 3.3.1  Results Food web structure  The distribution of body size and niche width in food webs varied considerably across the parameter space investigated (Figure 3.2). The size-structuring of predation led to a strong correspondence between body size and trophic position. In most simulations integer trophic positions (trophic levels) were occupied by specialists with narrow niches and body sizes close to multiples of the predator-prey size ratio d. These specialists had high densities due to their high rate of consumption of the trophic level below them, and were thus able to exclude species with similar body sizes and wider niches through both exploitative and interference competition. However, species with larger niche widths could often persist if their body sizes differed sufficiently from the specialists; generalists with intermediate body sizes (between d and 2d, or between 2d and 3d) experienced less competition and were able to consume prey from two or more trophic levels. When niche width was unconstrained (c = 0) some species evolved large body sizes (x > 9) and very wide niches (s > 5), giving them an almost flat utilization curve that allowed them to feed on all smaller species (and the basal resource) with low efficiency (Figure 3.2F). These species had relatively low trophic position despite their large size because much of their diet came from the basal resource. Food webs that evolved without constraints on niche width were highly variable in species trait composition and the number of specialist trophic levels. They also occasionally underwent evolutionary suicide (extinction of all species), which appeared to occur when large predators drove smaller species extinct and then went extinct themselves when they were unable to persist on the basal resource. The introduction of a cost to deviating from the optimum niche width reduced the variability in niche width, but when the cost was moderate (e.g. c = 0.5, Figure 3.2B,D) there was still marked variability in niche width that was associated with body size. When the optimum niche width was small, specialists and generalists coexisted with specialists at integer trophic levels and generalists between (producing a hump-shaped relationship between niche width and body size; Figure 3.2D). When the optimum was larger there was a trend toward increased niche width at higher body sizes and the absence of specialists at upper trophic levels (Figure 3.2B). As cost increased further niche widths were increasingly clustered around the optimum, with either a continuum of body sizes (at large s0 ; Figure 3.2A) or a linear food web with all species at trophic levels (at small s0 ; Figure 3.2C). The combination of high cost and a narrow optimum prevented 28  8  B  1.25  0.25  c =4  D  1.25  0.25  c = 0.5  6  0.50  0.75  *  1.00  0.50  0.75  *  1.00  1.25 c = 0.5  4 2 2  4  6  C  0  Body Size (x)  8  0.25  *  0.50  0.75  1.00  E  *  0.50  0.75  1.00  1.25  F  c =4 σx = 0.25 σs = 0.1  c =0  0  2  4  6  8  0.25  Body Size (x)  c =4 TP = 3 TP = 2 TP = 1  0  Body Size (x)  A  0.25  *  0.50  0.75  1.00  1.25  0  Niche Width (s)  1  2  3  4  5  6  7  8  Niche Width (s)  Figure 3.2: Representative food webs produced by the model under different parameter combinations. Each panel shows body size (x) vs. niche width (s), with symbol size indicating species trophic positions and lines connecting predators with prey that comprise 5% of their diet. Mutation size distributions were σx = 0.5 and σs = 0.2 for all panels aside from (E). The cost parameter c is shown in each panel, and the optimum niche width s0 is indicated with an asterisk (except where c = 0). Note that the placement of the basal resource on the x-axis is arbitrary, as it has no niche width.  29  evolution beyond the second trophic level when the distribution of mutations was small (Figure 3.2E). In most other cases, the structure of food webs was dictated by the foraging parameters c and s0 rather than the evolutionary parameters (σx and σs ). This variability in trait composition led to differences in the trophic structure of food webs simulated under different conditions. Neither species richness nor trophic height varied strongly with the parameters we investigated; most webs had 20-40 species and TPmax ≈ 3 (with a few exceptions described above). Despite these similarities in species richness and trophic height, food webs differed markedly in their distributions of trophic positions and interaction strengths, depending on the values of c and s0 . Mean omnivory generally declined with increasing cost (especially from c = 0 to c > 0), and increased with the optimum niche width.  3.3.2  Dynamics of species turnover  Species turnover, quantified as the proportion of mutations that led to a successful invasion of the food web, was influenced by each of the main parameters we varied (c, s0 and σx and σs ). Turnover generally decreased with increasing cost, increased with larger optimum niche width, and decreased with the size of trait mutations. Rates of food web assembly were driven largely by the size of trait mutations: for example, regardless of the values of c and s0 , food webs reached a second trophic level (TPmax ≥ 2.0) approximately 3× faster when mutation sizes were high (average ∼ 10, 000 vs. ∼ 30, 000 generations). Phylogenetic tree shape showed relationships with c and s0 (Figure 3.3). Gamma was usually negative (indicating an early accumulation of extant lineages) for any simulation with c > 0 (Figure 3.3 A-E), while c = 0 led to positive gamma values due to a large recent upturn in the LTT plots (Figure 3.3F). Gamma decreased with cost when s0 = 1 and increased slightly with cost when s0 = 0.5. Visual inspections of the LTT plots confirmed that Gamma was a useful descriptor of this aspect of tree shape (Figure 3.3). Gamma was correlated with species turnover (Spearman’s rank correlation: ρ = 0.44), indicating that patterns in phylogenetic tree shape were driven by differences in rates of turnover and community variability and invasibility. Variability in food webs evolved under different conditions led to emergent relationships between trophic structure and assembly dynamics (Figure 3.4). Across the parameter space explored here, mean omnivory was positively correlated with both species turnover (ρ = 0.70) and gamma (ρ = 0.57).  3.3.3  Evolutionary patterns of trophic position  The relative support for different evolutionary models of trophic position varied with trophic structure (Figure 3.4). The early burst model generally performed best in simulations with low levels of omnivory, while the BM or OU models were more often favoured in simulations with more omnivorous species. In other words, when species tend to occur at integer trophic levels, 30  50 20  30  40  c = 0.5 s0 = 1  B  10  Gamma = -5.56 Gamma = -1.7  c =4 s0 = 0.5  C  c = 0.5 s0 = 0.5  D  Gamma = -5.15  Gamma = -4.68  c =4 s0 = 0.5 σx = 0.25 σs = 0.1  E  c =0 s0 = 1  F  10  20  30  40  50 0  10  20  30  40  50 0  Species Richness Species Richness  Gamma = 5.15  Gamma = -1.1  0  Species Richness  c =4 s0 = 1  A  0  2.5 × 105  5 × 105 7.5 × 105 Generation  1 × 106 0  2.5 × 105  5 × 105 7.5 × 105 Generation  1 × 106  Figure 3.3: Representative examples of the dynamics of food web assembly under different parameter combinations. Values of c and s0 are indicated in each panel, and the mutation size distributions were σx = 0.5 and σs = 0.2 for all panels aside from (E). The solid line shows species richness at 1000-generation intervals, while the dashed line is a lineage-through-time (LTT) plot based on the phylogeny of extant species at the end of the simulation, with its shape described by the gamma statistic (see text). Symbols indicate when the food web first attained a trophic height (TPmax ) of 2.0 (inverted triangle) and 3.0 (diamond).  31  ● ●  ● ●  ●  ● ●● ● ● ●● ● ●●  ●  ● ●  ●  ●  ● ● ● ● ● ● ● ●● ● ● ● ● ● ●  ●  ●  ● ●  ●  ●  ●  ● ●  ●  ● ●  ●  ●  ● ● ● ●  ●●  ● ● ●  ●  ρ= 0.57  ●  ● ●  ●  ●  ●  ●  ● ●  ● ● ● ●  ●  ●  ●  ●  ●  ●  ●  −4 0.2  0.4 Mean Omnivory  0.6  0.0 0.0  0.2  0.4 Mean Omnivory  0.6  0.0  ● ●  ● ● ●● ● ●● ● ● ● ●●  ●  ●  ●  ρ= 0.73  ● ●  ●  ●  ●  −6  0.0  ●  ●  ●  0.0  ●  ● ● ●  ● ● ● ● ●  ρ= 0.62  ●  ●  0.2  ●  0.1  ●  ●  ●  ● ● ● ●  ● ● ●  ●  ●  ●  2 ●  1.0  6  ●  ●  ● ●  ●●● ● ●● ● ●● ●  −2  Species Turnover 0.2 0.3  ●  ● ●  ● ●● ● ●  ●  ● ●  ● ● ●  ●  ●  ρ= 0.70  ● ● ●  ●  ● ●  ●  ● ● ●  ●  ●  ●  ●  ● ●  ●  ●●  Akaike Weight (Early Burst) 0.4 0.6 0.8  ● ●  ● ● ●  ●  ●  4  ● ●  Gamma 0  0.4  ● ● ●  ●●  0.2  ● ●●●●●  ● ●  ●  ●● ●  ●  ● ● ●  ● ●●●  ●  ρ= −0.74  ● ●●● ●  0.4 Mean Omnivory  ●  0.6  Figure 3.4: Relationships between trophic structure, species turnover and evolutionary model fit. A measure of trophic structure, mean omnivory, is compared to the rate of species turnover (proportion of mutants that had positive growth rates and persisted in the community), the gamma statistic describing the temporal distribution of nodes in the phylogeny of extant species, and the relative support for the ‘early burst’ model of trophic position evolution. Symbols indicate values of the parameters c (circles: c = 0; triangles: c = 0.5; squares: c = 4) and s0 (open symbols: s0 = 0.5; filled symbols: s0 = 1). Lines are locally-weighted polynomial regressions (lowess) with a smoothing parameter of 0.75. In the right panel, the lowess lines for the support of all three models are shown (solid: EB; dashed: BM; dotted: OU). most of the divergence in trophic position is accounted for by early splits in the phylogeny, but when species are more omnivorous there is a greater degree of divergence between closely related species (Figure 3.5).  3.4  Discussion  Our simulations show how different forms of foraging trade-offs may influence both the structure and dynamics of food webs. With size-structured interactions, advantages to feeding on a wide or narrow range of prey sizes promote the assembly of food webs with higher or lower trophic complexity, respectively. These differences in foraging trade-offs along with the sizes of trait mutations also affect assembly dynamics, species turnover and even phylogenetic tree shape. Across a range of conditions these patterns lead to emergent relationships between food web structure and stability. We find that conditions favoring higher levels of omnivory also tend to increase the variability of food webs and species turnover through time. The structure of food webs produced by our model depends more on the foraging trade-off parameters – the optimum niche width and the cost to deviating from this optimum – than on the  32  Figure 3.5: Illustration of how trophic position maps to phylogeny under different conditions in the assembly model. Red lines connect the same species in the phylogenetic tree and in the food web. Trophic links in the food web connect predators to prey that comprise ≥ 1% of their diets. Left: a food web with discrete trophic levels (mean omnivory = 0.0001). Right: a food web with extensive omnivory (mean omnivory = 0.353). size of mutational steps. Thus, we expect the relationship between food web structure and stability to be similar whether new species arise from speciation or from invasion from a regional pool (see also L&L). However, in a few cases the dependence on evolutionary assembly constrained the development of complex food webs. Traits associated with omnivory in our simulations (large s and x intermediate between multiples of d) often provided a ‘bridge’ allowing the subsequent evolution of specialists at the next highest trophic level. Thus, for parameter combinations where these omnivory-associated traits have very low fitness (large c, small s0 ) and mutations are small (small σx and σs ), higher trophic levels establish very slowly or not at all over the course of a simulation (for example, with a σx of 0.2 only one in 1022 mutants will gain two body size units in a single step). Additionally, when costs are absent, there are often two types of species – small specialists and large generalists – that are so far apart in trait space that the specialists cannot be replaced if they become extinct. In these cases, migration of species from a regional pool could rescue the local food web following extinction of the small specialists. Our approach investigates how varying foraging trade-offs can lead to emergent relationships between food web structure and stability. This contrasts with previous theoretical investigations of omnivory-stability relationships, which have varied parameter values (e.g. attack rates) within fixed food web configurations (Holt and Polis, 1997; Diehl, 2003; Vandermeer, 2006). In our model, c and s0 influence the shape of the adaptive landscape – peaks may be steep and concentrated around traits suitable for specialist trophic levels, or more diffuse with a range of trait values able to invade and coexist. The higher variability and species turnover in food webs with omnivory seems to result from the flattened adaptive landscape, which increases the proportion of trait space that can maintain viable populations. Additionally, food webs with omnivores have fewer dominant, strongly interacting species, which may make them more 33  susceptible to invasion (Case, 1990). Thus, a positive relationship between omnivory and temporal stability emerges from variation in environmental conditions (trade-off form) and speciation-extinction dynamics. The analysis of evolutionary model fits leads to a testable prediction of this model. Namely, we can predict that trophic position is likely to fit an ‘early burst’ model of rapid divergence followed by stasis only in clades that occupy stable and discrete trophic levels. This is because, as described above, once a lineage has established in a discrete trophic niche, there should be a low probability of another species evolving traits that allow it to exploit the same niche fast enough to avoid competitive exclusion by the resident species. In contrast, when discrete trophic niches do not exist, or are not temporally stable, there is a greater chance of species displacing one another and models of continuing trait change (BM or OU) tend to fit better. As the range of trophic niche space that can be explored is constrained by practical limits (the logical lower bound of 1.0 and, in the model, an upper bound of 3-4 based on energetics), this should lead to the improved fit of the OU model, in which recurrent trait evolution occurs within a confined range of trait space. Previous studies of food webs have identified features of food webs that vary among ecosystem types, including incidence of omnivory, trophic height, degree of size-structuring and distribution of biomass among trophic groups (Schoener, 1989; Shurin et al., 2006; Thompson et al., 2007; Vander Zanden and Fetzer, 2007). The relationship between omnivory and stability in our model depends on variation in environmental features that determine the form of foraging trade-offs. One feature of real ecosystems that might influence such trade-offs is the structural complexity (grain size) of the environment (Ritchie and Olff, 1999). If environmental heterogeneity allows size-based partitioning of habitat, consumers may be penalized for foraging on different prey sizes due to costs of maintaining foraging tactics at multiple spatial scales (i.e. low s0 and c > 0 in our model). Based on our results we would predict that such a community would have relatively low levels of omnivory and high temporal stability. On the other hand, if trophic interactions occur in a well-mixed environment it may be costly to be size-selective (i.e. high s0 and c > 0), leading to more omnivory and potentially greater species turnover. An examination of empirical body size data suggests that prey-size generality (the standard deviation of body sizes of a consumers prey) varies across ecosystems (Ingram et al., 2009), appearing to be highest in freshwater, intermediate in marine, and lowest in terrestrial systems. These patterns suggest that ecosystems vary in prey size generality (roughly corresponding to niche width in our model), and raise the possibility that trade-offs to resource use also vary across systems. Future comparisons across and within ecosystems should allow a more detailed understanding of variation in dietary niche widths, and the implications of this variation for food web structure and dynamics. In our model, a single adaptively radiating lineage diversifies to fill a whole food web, allowing the examination of phylogenetic tree shape in addition to structural and dynamic 34  features of food webs. We found that the shape of reconstructed lineage-through-time plots depended on the values of the ecological trade-off parameters. This pattern results from the differences in dynamic stability already described: the more species turn over, the fewer lineages will persist for long periods of time and the more closely extant species will be related (Pybus and Harvey, 2000). When turnover is lower lineage accumulation occurs earlier and the rate of (successful) speciation slows down as the community approaches equilibrium. Both of these situations correspond to early niche-filling during adaptive radiation (Gavrilets and Vose, 2005), while the variability in tree shape largely comes from whether subsequent turnover overwrites the early history of the clade by removing older lineages. Our analysis of phylogenetic tree shape contributes to a growing interest in interpreting variation in real phylogenies with regard to ecological interactions (Phillimore and Price, 2008; Rabosky and Lovette, 2008; McPeek, 2008). In a recent metacommunity simulation study, McPeek (2008) showed that clades had negative gamma when speciation was associated with ecological divergence, and positive gamma when speciation generated ecologically equivalent species. While all of our speciation events generated differences in trait values, our results are consistent with these findings. Our food webs with high omnivory included more weakly interacting species that likely had very small differences in fitness, similar to the non-ecological speciation model used by McPeek (2008). The implication that gamma can be affected by both the mode of speciation (ecological and non-ecological) and the specific context of ecological speciation (different trade-off structures in our model) should lead to more detailed investigations into variability in empirical phylogenetic tree shape. Some organisms, especially fishes such as cichlids and salmonids, appear to be prone to diversifying at multiple trophic levels during adaptive radiations (Sk´ulason and Smith, 1995; Schluter, 2000). When diversification occurs both within and between trophic levels, our results suggest that features of the environment that influence trophic niche evolution may also influence phylogenetic tree shape in adaptively radiating lineages. Food webs with omnivores are less stable in our model, in terms of species turnover, than those with all species at integer trophic positions. How can this result be reconciled with empirical observations that omnivory is ubiquitous and with theoretical findings that omnivory should often be stabilizing? A possible explanation is that while turnover increases with omnivory, other components of food web stability – such as permanence – may not decrease. In our model, global extinctions of all species rarely occurred; rather, food webs tended to reach a dynamically stable configuration with little structural change despite constant turnover. Omnivory may also contribute to stability by allowing greater resilience to perturbations (such as species loss or changes in nutrient levels). Our model features a constant environment (e.g. nutrient input), and apart from speciation all dynamics are deterministic. Incorporating environmental stochasticity and disturbance would be a useful extension to investigate whether 35  omnivory confers stability in a variable environment. Alternatively, omnivory may be common in food webs despite being associated with instability if food webs are not stable entities over the time scales considered here. Much work remains to elucidate the various direct and indirect associations between trophic structure and stability in real food webs. Our model provides a framework for investigating the simultaneous evolution of body size and niche width in food webs, and the resulting development of complex trophic structure. We find that trade-offs to resource use may dictate the amount of omnivory that occurs in food webs, and lead to emergent relationships between omnivory and species turnover, variability and even phylogenetic tree shape. By exploring how species interactions and foraging constraints contribute to community-wide patterns, we may gain a more thorough understanding of the relationships between community assembly, trait evolution and food web structure.  36  Chapter 4  Tempo and mode of trophic position evolution in northeast Pacific rockfish (Sebastes) 4.1  Introduction  The diversity and structure of food webs has important implications for their stability and function. Most of the research linking diversity to ecosystem function has focused on ‘horizontal’ diversity – the number of species at a single trophic level (Cardinale et al., 2006). ‘Vertical’ diversity – the number of trophic levels and the distribution of species among them – can have equally important effects on ecological processes (Duffy et al., 2007; Srivastava and Bell, 2009), making it important that we understand the factors that influence the vertical structure of food webs. Vertical structure of food webs is traditionally described by organizing species into discrete trophic levels, but evidence for widespread omnivory in real food webs (Polis, 1991; Goldwasser and Roughgarden, 1993) has led to the adoption of a continuous measure of trophic level. Trophic position measures the average number of trophic steps separating a species from the base of the food web, and can be estimated using stable isotope tracers or food web link data (Levine, 1980; Vander Zanden et al., 1997; Post, 2002b). The vertical structure of food webs is affected by contemporary ecological processes such as dispersal and adaptive foraging, as well as by evolutionary processes such as speciation and trait evolution. A consideration of the evolutionary processes that underlie divergence in trophic position among species may therefore be required to both understand present patterns of trophic structure and to predict future evolutionary responses to species loss. Many groups of organisms evolve little or no diversity in trophic position over long time periods, but substantial vertical diversification does occur in many other clades, including some 37  birds, reptiles, insects and fish (R¨uber et al., 1999; Sato et al., 1999; Jonsson, 2001; Fagan et al., 2002; Espinoza et al., 2004). Trophic position may evolve due to selection acting on other traits such as body size, or due to adaptation to new habitats with longer or shorter food chains (Matthews et al., 2010). Alternatively, the evolution of trophic position may be driven by fitness advantages associated with feeding on plants, herbivores or carnivores, such as a general increase in resource quality and decrease in quantity with trophic position (Fagan et al., 2002; Hendrixson et al., 2007). In this case, it may be appropriate to think of trophic position as being a more direct target of natural selection. Evolutionary diversification in trophic position has important implications for the structure and stability of food webs, due to effects of predator presence and diversity on lower trophic levels. Loss of top marine predators can lead to cascading effects across lower trophic levels that alter primary productivity (Frank et al., 2005; Myers et al., 2007; Heithaus et al., 2008). There is some evidence that fisheries selectively target large species with high trophic levels (Pauly et al., Pauly et al.; but see Essington et al., 2006; Branch et al., 2010) which can result in the ecological extinction (Frank et al., 2005) or decreased body size (Darimont et al., 2009) of these important top predators. To begin to understand the long-term consequences of top predator depletion or extinction, it may be useful to study the tempo and mode of trophic position evolution in lineages undergoing adaptive radiation. For example, if a relaxation of fishing pressure following top predator extinction led to natural selection for higher trophic position, we may ask how readily such an evolutionary replacement could occur? If trophic position evolves rapidly and is weakly constrained, the evolutionary replacement of top predators may be more likely than if the evolution of trophic position is subject to stronger constraints. One approach to infer evolutionary processes that affect the evolution of a trait such as trophic position is to compare the distribution of trait values among extant species to a phylogeny describing their evolutionary relationships. By comparing the fit of alternate models, we may be able to identify ecological and evolutionary processes that can explain the patterns we see today. Despite the increasing availability of comparative methods, very few studies have yet investigated the evolution of trophic position in such a framework. One study calculated a positional index related to trophic position for 116 taxonomically diverse fish species in an empirical marine food web (Rezende et al., 2009), and another examined stable isotopes as a proxy for trophic position in Lake Tanganyika cichlids (Wagner et al., 2009). Both studies demonstrated that trophic position showed phylogenetic signal (meaning phylogenetic relatedness and similarity in trophic position are correlated), but no study has yet evaluated multiple macroevolutionary models to investigate the dynamics of trophic position in an adaptively radiating lineage. Before investigating the fit of evolutionary models to trophic position data on a phylogeny, it is desirable to have hypotheses for what patterns we might expect under different conditions. There is little if any theory to guide these hypotheses, but the development of evolutionary 38  assembly models of food webs may be a useful starting point. These models allow vertical structure in food webs to emerge through speciation and the evolution of traits that affect trophic interactions (Caldarelli et al., 1998; Loeuille and Loreau, 2005; Ingram et al., 2009). One prediction to arise from one of these models (see Chapter 3) is that the existence of discrete trophic levels should produce a pattern of an ‘early burst’ of trait evolution (i.e., most trait variation is accounted for by differences between subclades that diverged early in a clade’s history). In contrast, when species in the clade are highly omnivorous, the pattern may be more likely to fit a model of recurrent trait evolution within some constrained trait space. These predictions arise from variation in the strength of foraging trade-offs in size-structured food webs, where strong trade-offs reduce the fitness of omnivores and thereby decrease the chance of evolution between trophic levels. Rockfish in the genus Sebastes exhibit substantial among-species variability in adult trophic position, making this a promising group in which to directly study the evolution of trophic position. Rockfish have diversified into over 100 species since the mid-Miocene (ca. 8 m.y.a.), approximately 70 of which occur in the northeast Pacific (NEP) between Alaska and Baja California (Hyde and Vetter, 2007; Love et al., 2002). My previous work on this group has indicated that measures of trophic position based on stable isotopes show a strong negative relationship with gill raker number and length, and a weaker positive relationship with body size (Ingram and Shurin, 2009; Ingram, 2011). I previously showed that the evolution of trophic morphology does not appear to be associated with speciation events (Ingram, 2011), though trophic diversity may play a role in species coexistence (Hallacher and Roberts, 1985; Ingram and Shurin, 2009). Here, I use higher quality isotopic baseline data to generate improved estimates of trophic position for 32 rockfish species, and develop a predictive model to estimate the trophic position of unmeasured species from their morphology. I estimate the rate of trophic position evolution and fit evolutionary models to identify classes of processes that may account for the present-day distribution of trophic positions. Using these inferred model fits, I also conduct evolutionary simulations to explore the possibility of a top predator evolving in a group such as Sebastes under different evolutionary scenarios.  4.2 4.2.1  Materials and methods Stable isotope analysis  I carried out stable isotope analysis on a total of 169 individual rockfish, representing 32 species, in 2006 and 2007. Sample sizes ranged from 1 to 17 adult individuals of each species, with a median of 5 (sample sizes and sources of specimens for each species are given in Appendix A). These sample sizes are generally sufficient for comparative analyses when variation among 39  species is high relative to within species (Harmon and Losos, 2005). Most samples used for isotope analysis were obtained from recreational fishing boats. Fish were either sampled while fresh the day of capture, or were frozen until processing. 116 individuals from 27 species were sampled in the Santa Barbara Channel in 2007, while 53 individuals from 9 species were sampled in Barkley Sound, British Columbia in 2006 and 2007 (4 species were represented in both regions). I did not include isotope data from 12 additional species that I used in previous studies (Ingram and Shurin, 2009; Ingram, 2011), because they were sampled from a wide range of depths (150-1000 m) from an area (the continental slope off Haida Gwaii, British Columbia) lacking suitable baseline data. A small sample of white dorsal muscle tissue from behind the operculum of each rockfish was frozen and then dried at 60 ◦ C for 48 h. Each sample was then ground to powder with an electronic amalgamator, and 1.0 mg was packaged into a tin capsule. Stable isotope analysis of nitrogen and carbon was conducted at the University of California, Davis Stable Isotope Facility using a PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). Isotope ratios are expressed in the conventional delta notation (δ 15 N and δ 13 C), as ratios relative to a standard (air for nitrogen, and Vienna Pee Dee Belemnite for carbon). Measurement error based on the standard deviation of replicated analyses of standards was 0.12  for δ 15 N and 0.04  for δ 13 C. Lipid synthesis  reduces δ 13 C in tissues, and variation in lipid content among consumers can bias estimates of carbon contributions to consumer’s diets when the C:N ratio (a proxy for lipid content) is greater than 3.5 (Post et al., 2007), as is the case in these samples. I therefore calculated lipid-corrected δ 13 C values as δ 13 C = δ 13 Craw − 3.32+ 0.99 × C : N (Post et al., 2007), resulting in a small adjustment of typically less than + 0.5 . I did not extract lipids directly because lipid extraction has an undesirable effect on the δ 15 N of rockfish tissue (Ingram et al., 2007). In order to convert isotope ratios to ecologically meaningful measures of trophic position and carbon source usage, it is necessary to calibrate them to baseline isotope values of organisms of known trophic position (Post, 2002b). In aquatic systems, benthic producers typically have higher (less negative) δ 13 C than pelagic producers, allowing the relative reliance of an organism on benthic vs. pelagic carbon to be accounted for in estimating its trophic position. I used particulate organic matter (POM) as a pelagic baseline, and the dominant habitat-forming kelp Macrocystis integrifola as a benthic baseline (Figure 4.1). Both POM and kelp were sampled across a wide spatial area in both Barkley Sound (kelp, n = 49; POM, n = 48; R. Markel, unpub. data) and the Santa Barbara Channel (kelp, n = 23; POM, n = 27; A. Salomon, unpub data). These baseline organisms are unlikely to perfectly capture the isotopic values at the base of the food web. POM contains a mix of phytoplankton and other particles that may be more enriched in δ 13 C than pure phytoplankton, while M. integrifola is one of many benthic producers that vary in δ 13 C. These sources of error increase uncertainty in the reliance of organisms on 40  ●  Santa Barbara Channel Barkley Sound  9  δ15N 12  15  18  ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ●●●● ●● ● ● ●● ● ● ● ●● ● ● ● ●●● ●● ● ●● ●● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ●● ●● ● ● ●●●● ●●●● ● ● ● ● ● ●●●● ● ● ● ● ● ● ● ● ●●● ●● ●  ●  6  ● ●  Macrocystis POM −22  −20  −18  −16  −14  −12  δ C 13  Figure 4.1: Raw δ 15 N and δ 13 C data for rockfish from the Santa Barbara Channel (circles) and Barkley Sound (squares), and mean and standard deviations for isotopic baselines (Macrocystis kelp and particular organic matter) from each region. Filled circles indicate species collected from deeper (200 m) water, and the adjusted δ 15 N pelagic baseline. benthic vs. pelagic carbon, but trophic position estimates are generally not highly sensitive to uncertainty in carbon baselines (Post, 2002b). A greater concern for estimating trophic position is the tendency for δ 15 NPOM to increase with depth in marine systems (Mintenbeck et al., 2007; Ostrom et al., 1997), which I could not account for directly in the absence of POM samples from deeper water. Most rockfish were caught in shallow (≤ 60 m) water, but a subset of the Santa Barbara Channel fish were caught at ∼ 200 m, and tended to have elevated δ 15 N values compared to fish of similar size and trophic morphology from shallow water (Figure 4.1). I thus added 1.5  to the δ 15 NPOM baseline for these deeper-caught fish, in accordance with typical 41  published enrichments in δ 15 N over this depth range (Mintenbeck et al., 2007; Ostrom et al., 1997). This adjustment effectively removed the size- and morphology-independent elevation of trophic position with depth. In typical calculations of the trophic positions of aquatic consumers, one first estimates the contribution of benthic carbon to the diet (in this case the proportion of kelp-derived carbon; PKDC) as PKDC =  (δ 13 C − CP ) , (CB − CP )  (4.1)  where CP and CB are the δ 13 C values of the pelagic (POM) and benthic (kelp) baselines, respectively (Post, 2002b). Trophic position is then estimated as TP = λ +  (δ 15 N − (PKDC × NB + (1 − PKDC)NP )) , ∆N  (4.2)  where ∆N is the mean per-trophic step change in δ 15 N, NP and NB are the baseline δ 15 N values, and λ is the trophic position of the baseline organism, in this case 1.0 (Post, 2002b). These formulae assume that the mean per-trophic step change in δ 13 C (∆C ) is zero, whereas meta-analyses tend to show an average ∆C of ∼0.5  (Vanderklift and Ponsard, 2003). If  ∆C = 0, PKDC and TP can be estimated from an iterative routine (Post, 2002b), but here I provide exact equations for these quantities. The following formulae can be used when ∆C = 0, and are derived from the intersection point of two lines in isotopic space. One line connects the benthic and pelagic baselines, and the other passes through the consumer’s δ 15 N and δ 13 Cand has slope ∆N /∆C . PKDC =  TP =  ∆N (CP − δ 13 C) + ∆C (δ 15 N − NP ) ∆N (CP − CB ) + ∆C (NB − NP )  (4.3)  CP (δ 15 N + λ ∆N − NB ) − CB (δ 15 N + λ ∆N − NP ) + (λ ∆C + δ 13 C)(NB − NP ) ∆N (CP − CB ) + ∆C (NB − NP )  I estimated TP and PKDC for each individual using values of ∆N = 3.4  (4.4)  and ∆C = 0.5  (Vanderklift and Ponsard, 2003; Post, 2002b) , and calculated the mean adult trophic position of each species.  4.2.2  Analysis of relationships between trophic position and morphology  I used a dated molecular phylogeny containing the 66 species in my dataset, produced using 7 mitochondrial and 2 nuclear genes (Hyde and Vetter, 2007) in the program BEAST (Drummond and Rambaut, 2007) (for details of the tree reconstruction, see Chapter 2 and Ingram, 2011). I 42  stored 100 trees from the posterior distribution to confirm that my results were robust to phylogenetic uncertainty. I pruned the tree to include only species found in the NEP (Figure 2.2). I compiled morphological data from 543 adult individuals from 66 Sebastes species from the NEP (for details see also Ingram and Shurin, 2009; Ingram, 2011). Traits included species’ maximum total length (TL) and gill raker number, as well as log-transformed and size-corrected gill raker length, eye width, body depth, lower jaw length and pectoral fin length (see Chapter 2 for details). I developed a multiple regression model to predict species’ mean trophic positions from their morphology. I used phylogenetic generalized least squares (PGLS) to model trophic position as a function of these seven traits, with the covariance matrix among observations derived from the phylogeny. I simultaneously estimated the λsig parameter (Freckleton et al., 2002) that describes the degree of phylogenetic signal of the predictor variables and the residual variation in the response variable. I used stepwise AIC to identify a minimal adequate model by sequentially deleting terms until the AIC score stopped improving. I used the coefficients of this simplified model to predict TP from morphology for the 34 species without direct estimates of TP (Garland Jr and Ives, 2000). These predictions did not take into account the phylogenetic position of unmeasured species, as there were too many missing data to properly use the available methods. This decision may increase prediction intervals, but does not bias predicted values of a response variable (Garland Jr and Ives, 2000).  4.2.3  Evolutionary analysis of trophic position  I used maximum likelihood to fit a number of candidate evolutionary models for trophic position. The most commonly used macroevolutionary model is one in which the variance in a trait accumulates linearly over time. This ‘Brownian Motion’ (BM) model can describe evolution by genetic drift, or adaptive evolution where the optimum trait value fluctuates randomly over time (Felsenstein, 1985). This model has two parameters, the ancestral state at the root of the tree, expressed as the phylogenetically weighted mean trait value x, ¯ and the Brownian rate parameter σ 2 , which describes the rate at which taxa diverge from each other over time. The rate estimated by maximum likelihood is slightly downwardly biased, but the difference should be trivial for a tree of this size (O’Meara et al., 2006). Under some verbal models of adaptive radiation, the rate of trait evolution is not constant as assumed under BM. For example, in a classic view of adaptive radiation, a rapid early filling of niche space is followed by evolutionary stasis (Simpson, 1953). I used the ‘Early Burst’ (EB) model (Harmon et al., 2010), in which an initial rate (σ02 ) declines exponentially over time at rate a. The patterns that result from this model are comparable to other models where the rate of trait evolution slows over time (Harmon et al., 2010). Alternatively, evolution by BM may fail to accurately describe evolutionary processes 43  because it assumes that disparity continues to increase indefinitely. This prediction is inconsistent with widescale findings that disparity is not strongly dependent on clade age, meaning older clades are inferred to have much lower evolutionary rates (Harmon et al., 2010). BM may thus underestimate short term evolutionary divergence and overestimate divergence over longer timescales. The Ornstein-Uhlenbeck (OU) ‘rubber band’ model modifies evolution by BM by including a tendency for trait values to return to an ‘optimum’ value with rate described by the parameter α (Hansen, 1997). This model was initially used to describe stabilizing selection around a single selective peak, but may also provide a reasonable description of the data when species adapt to an optimum that moves within a constrained space, or when trait space itself is bounded (Estes and Arnold, 2007). Another possible model is ‘speciational’ evolution, where much or all of the change in trait values is associated with speciation events. As my other work has indicated that there is no evidence for speciational evolution of rockfish trophic position (Chapter 2; Ingram, 2011), I do not consider this model any further here. I used likelihood estimation to fit each of the three models (BM, EB and OU), finding the parameter values that maximized the probability of observing the data given the phylogeny. The analytic approach behind these analyses is described in Chapter 2 and by Harmon et al. (2010). I fit the three models to the 66-species dataset combining predicted and isotope-derived trophic position values, as well as to the 32-species isotope dataset alone. To evaluate whether patterns are similar for body size and size-independent morphology, I also fit log-transformed maximum TL and gill raker number using the same models (note that these traits are not independent of the 66-species trophic position dataset). I accounted for measurement error in each trait by lengthening the terminal branches of the phylogeny. I used error estimates based on the average intraspecific standard deviations in my dataset, converted to standard errors assuming sample sizes of 5 (SE = 0.045 for trophic position, 0.042 for total length, and 0.496 for gill raker number). I compared the models using Akaike’s Information Criterion (corrected for small sample size: AICc ), which balances fit against model complexity by penalizing models with more parameters. The BM model has two parameters, while the remaining models have three each. I also calculated Akaike weights as a measure of the relative support for each model out of the set of models considered (Wagenmakers and Farrell, 2004). Finally, I conducted simple simulations that focussed on the best-fitting model (OU). I assumed that evolution continues to follow this OU process with the same values of α and σ 2 , but that evolution is toward a new optimum of TP = 5, comparable to many marine top fish predators (Thompson et al., 2007; Vander Zanden and Fetzer, 2007). This corresponds to a scenario in which the loss of top predators results in ecological opportunity favouring increased trophic position in Sebastes. I simulated the distribution of waiting times until a lineage starting 44  Table 4.1: Regression model for predicting trophic position from morphology Parameter Intercept Gill Raker Number Lower Jaw Length Body Depth Pectoral Fin Length Maximum Total Length  Estimate 3.79 -0.04 0.91 -1.15 1.20 0.34  Standard Error 0.63 0.009 0.38 0.31 0.61 0.11  t 6.05 -4.85 2.42 -3.70 1.94 2.95  p-value <0.0001 <0.0001 0.022 0.001 0.063 0.007  The regression model was identified using generalized least squares with covariance structure based on the phylogenetic tree. Stepwise AIC was used to eliminate terms (gill raker length and eye width) until the AIC stopped improving. at a trophic position of 4 (toward the upper end of the observed range of rockfish species) first reached the new optimum. I also compared the waiting times under this peak-shift scenario to the best-fit OU and BM models. For these simulations, I used a discrete-time approximation to the OU process with a step size ∆t of 0.05, where the change at each step is ∆tα(θ − x) + ε, where x is the character state at the previous step and ε is a random normal deviate with variance ∆tσ 2  4.3  Results  Isotope-derived estimates of Sebastes species mean trophic positions (TP) ranged from 3.18 (calico rockfish, S. dallii) to 4.25 (starry rockfish, S. constellatus), while the proportion of kelp-derived carbon (PKDC) ranged from 0.07 (swordspine rockfish, S. ensifer) to 0.64 (kelp rockfish, S. atrovirens). Within-species variation was relatively low (species identity accounted for 77% of variation in individual trophic position), so sample sizes of ∼5 individuals per species were appropriate for comparative analyses (Harmon and Losos, 2005). TP was best predicted by a linear model including gill raker number, jaw length, body depth, pectoral fin length, and maximum total length (Table 4.1). As shown previously (Ingram and Shurin, 2009; Ingram, 2011), trophic position was negatively related to gill raker number and positively related to maximum size. Additionally, trophic position was positively associated with jaw length and pectoral fin length, and negatively related to body depth. Under the BM model, the maximum likelihood evolutionary rate is 0.029. This value can be interpreted as the expected increase in variance (in units of trophic positions) per million years. Of the three evolutionary models, the constrained (OU) model was slightly preferred over the BM or EB models based on Akaike weights (Table 4.2) for the isotopic estimates of trophic position alone (wA = 0.58). For the expanded datasets that included trophic position predicted by morphology OU was strongly preferred over the other models (wA > 0.99). OU also provided the best fit for the key morphological traits: gill raker number (wA = 0.81) and maximum total length 45  4.4  SY  GB PI MX  PR  BO  YE  GP  TF  3.8  BR BA  BY GO OL VE  CO QB  GS  3.6  Trophic Position  4.0  4.2  FL  SL  SS  3.4  HB  YT  CP  WI CM  CA  KE CH  BL  3.2  BU  CL  0.0  0.2  0.4  0.6  0.8  Proportion Kelp−Derived Carbon Figure 4.2: Estimates of trophic position and PKDC for 32 Sebastes species (2-letter species codes as presented in Figure 4.3). Species means are presented, along with standard deviations where n > 1 (wA = 0.61). These results were robust to phylogenetic uncertainty: of the 100 trees sampled from the posterior distribution, OU was the best fit to all 100 for the expanded trophic position dataset, and for 96 of the isotope-only dataset (BM fit best in the other 4). OU also fit best for all 100 trees for gill raker number, and for 98 trees for maximum total length (again, BM fit the other two best). There was no evidence for a slowdown in trophic position evolution, as the MLE of the a parameter was 0 (corresponding to evolution by BM). These results were also consistent using two other reduced datasets. First, I ran the analyses excluding the grass rockfish, S. rastrelliger. This relatively large species has very few, short gill rakers, which caused it to have the highest predicted trophic position. This probably represents a 46  Trophic Position S.macdonaldi S.ruberrimus S.melanostomus S.aurora S.phillipsi S.gilli S.diploproa S.melanosema S.semicinctus S.saxicolaS S.atrovirens S.carnatus S.chrysomelas S.maliger S.caurinus S.nebulosus S.dallii S.rastrelliger S.auriculatus S.elongatus S.paucispinis S.jordani S.goodei S.mystinus S.entomelas S.flavidus S.serranoides S.melanops S.levis S.babcocki S.nigrocinctus S.serriceps S.rubrivinctus S.ensifer S.eos S.chlorostictus S.rosenblatti S.helvomaculatus S.simulator S.rosaceus S.umbrosus S.lentiginosus S.notius S.constellatus S.ovalis S.hopkinsi S.moseri S.rufinanus S.rufus S.pinniger S.miniatus1 S.crameri S.reedi S.polyspinis S.ciliatus S.variabilis S.alutus S.zacentrus S.emphaeus S.variegatus S.wilsoni S.proriger S.borealis S.brevispinis S.aleutianus S.melanostictus  *  *  MX YE  ● ● ● ●  CM  ● ● ● ●  HB SL KE GO BY QB CO CH CL  ● ● ● ● ● ● ● ● ● ●  BR GS BO  ● ● ● ●  CP BU WI YT OL BL  ● ● ● ● ● ● ● ● ●  TF FL SS PI GP GB  ● ● ● ● ● ● ●  PR  ● ● ● ● ●  SY  ● ● ● ● ●  BA CA VE  ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●  3.0  3.5  4.0  4.5  Figure 4.3: Sebastes phylogeny used for the evolutionary analysis of trophic position. The tree is identical to the tree in Chapter 2, but with the species from outside the northeast Pacific removed. The trophic position values displayed (in units of trophic levels) are either estimated directly from stable isotope data (filled symbols) or indirectly from morphology (open symbols). Species codes for species with isotope data can be matched to Figure 4.2. Asterisks highlight one species and one clade discussed in the Results. 47  Table 4.2: Results of evolutionary model fitting of trophic position and morphological traits Model  BM EB  OU  BM EB OU  Parameter x¯ σ2 L σ02 a L σ2 α L wA wA wA  Trophic Position (n = 32) 3.69 0.019 -2.68 0.019 0 -2.68 0.037 0.237 -0.89 0.329 0.096 0.575  Trophic Position (n = 66) 3.73 0.029 -16.15 0.029 0 -16.15 0.071 0.390 -10.42 0.010 0.003 0.987  Gill Raker Number 32.8 4.69 -182.8 4.69 0 -182.8 8.46 0.222 -180.0 0.141 0.047 0.812  Max. Total Length 4.05 0.044 -28.39 0.044 0 -28.39 0.069 0.162 -26.57 0.295 0.098 0.607  Model results are shown for both 32 species with trophic position estimated from stable isotopes, and for the expanded 66-species dataset that includes trophic positions predicted from morphology. x¯ is the estimated root state (and the estimated optimum for the OU model), σ 2 is the Brownian rate parameter, a is the slowdown parameter in the EB model, α is the constraint parameter in the OU model, and L is the log-likelihood. The bottom three rows show the support for each model using Akaike weights (wA ) based on AICc , with the preferred model highlighted in bold. failure of the predictive model, as dietary information on this species suggests that it is feeds mainly on hard-bodied benthic invertebrates, in contrast to other large species with few gill rakers which are more piscivorous. Predicted trophic position was still clearly best fit by the OU model with this species excluded (wA = 0.98). I also repeated the analysis of predicted trophic position after excluding the species in a clade that is sister to the rest of the NEP Sebastes, and includes mostly deep-living species as well as some species whose range extends to the western Pacific. As this clade is not represented by isotopic data, it is possible that their trophic position is not reliably predicted using the model derived for species in the larger clade. However, the analysis excluding these 15 species yields very similar results (wA = 0.96 in favor of the OU model). The OU model often fits empirical datasets well, but that does not necessarily mean that trait evolution is in fact governed by stabilizing selection around a single broad adaptive peak. Other models may share the OU model’s key feature of recurrent evolution within a constrained trait space. For example, a model of bounded Brownian motion (BBM) with hard bounds on trait space shares these features and is a simple and viable alternative to the OU model (Ackerly, 2009; Revell et al., 2008). Trophic position has a hard lower bound (2, for heterotrophs) and is likely constrained to a maximum of 5.5-6 in marine ecosystems due to trophic transfer efficiencies and ecosystem size (Pauly and Christensen, 1995; Post, 2002a; Vander Zanden and Fetzer, 2007), so  48  it might be particularly likely to fit this bounded model. Simple likelihood methods for inferring bounded Brownian motion have not been published, but a formulation using partial differential equations that integrates backward in time is currently in development (R. G. FitzJohn and T. Ingram, unpub. data). I used code from the R package ‘diversitree’ (FitzJohn, 2010) to infer the parameters of BBM (σ 2 , lower bound, and upper bound). Based on a preliminary comparison of BBM to the OU model, OU provides a better fit for trophic position (wA = 0.72 for isotope data only and wA = 0.99 for the 66-species dataset) and total length (wA = 0.68), but fits less well than BBM for gill raker number (wA = 0.34). The evolution of rockfish trophic position thus is not necessarily constrained by hard bounds on the trait space that can be explored. If we tentatively take the OU fit as reflecting stabilizing selection around an optimum, we can use the estimates of the strength of the constraining force α to predict the rate of approach to a new optimum following a shift in the adaptive peak (Hansen, 1997). Focusing first on the deterministic part of the OU model, a value of α = 0.39 implies a time to evolve halfway to a new optimum of log(2)/α = 1.78 million years (m.y.), or 2.92 m.y. based on the lower value of α estimated from the 32-species isotopic dataset. The simulations that incorporate both the stochastic and deterministic parts of the OU model give some idea of the probability and rate of evolution to a new optimum. These simulations predict that if α and σ 2 do not change as a species with a starting trophic position of 4.0 adapts to a new optimum at 5.0, it will reach this new optimum in a median of 4.85 m.y. The distribution of waiting times ranged from less than 2 to greater than 10 m.y., by which time the new optimum had been reached in 92% of simulations (Figure 4.4). For comparison, a trophic position of 5.0 was only reached by 10 m.y. in 2.3% of the simulations under the estimated parameters of the BM model, and was never reached by 10 m.y. under the OU model without a peak shift.  4.4 4.4.1  Discussion Variation in trophic position in Sebastes  These analyses reveal that rockfish vary over almost 1.5 trophic levels, from a trophic position close to 3 for species consuming mainly herbivorous animals to near 4.5 for largely piscivorous species. While quantitative estimates of trophic position are not widely available for many taxa, this is likely to be a relatively high degree of trophic position variation for a group of closely related species. Results were generally consistent with available diet data based on stomach content analysis; estimates of trophic position were generally low for species known to consume primarily plankton and higher for species known to be more piscivorous (Love et al., 2002). Following baseline corrections, trophic position was not strongly associated with the proportion of kelp-derived carbon estimated in species’ diets: species with high trophic position can be either relatively kelp-associated (e.g. yelloweye rockfish, S. ruberrimus) or not (e.g. starry 49  0  2  4  6  8  10  >10  Time to Trophic Position = 5 (m.y.)  Figure 4.4: Truncated distribution of waiting times until a Sebastes species with a trophic position of 4.0 is predicted to evolve a new trophic position of 5.0. For this set of simulations, I used the parameters of the fitted OU model (α = 0.39 and σ 2 = 0.071) and assumed adaptation to a new ‘optimum’ trophic position of 5.0. rockfish, S. constellatus), as can species with lower trophic position (e.g. kelp rockfish, S. atrovirens vs. swordspine rockfish, S. ensifer). There was no apparent tendency for trophic positions to cluster around integer values, suggesting that rockfish generally do not occupy discrete trophic levels (Figure 4.2). I also confirmed previously documented relationships between rockfish morphology and trophic position (Ingram and Shurin, 2009; Ingram, 2011). Large body sizes and mouth (or jaw) sizes are naturally associated with trophic position in many size-structured fish communities, though individual size is often a better predictor than mean species size (Jennings et al., 2001; Layman et al., 2005). Gill raker number and length are related to use of small prey such as zooplankton in many fishes (Drenner et al., 1984; Palkovacs and Post, 2009). Small zooplankton appear to be the lowest trophic position prey used by rockfish, so the most planktivorous species tend to have the lowest trophic position. The relationships between trophic position and both pectoral fin length and body depth have not previously documented, and their repeatability and functional significance remain to be determined.  4.4.2  Interpretation of evolutionary models  Rockfish trophic position and trophic morphology were best fit by the Ornstein-Uhlenbeck (OU) model. There are two key features of the OU model that are true regardless of the processes that underlie its relatively good fit. First, unlike under Brownian motion, the amount of trait space that is explored through evolution is restricted in some way, such that variance among species does 50  not increase indefinitely with time. Second, evolutionary change continues apace throughout the clade’s history, unlike the early burst model in which the rate of trait evolution slows once trait space is full. As a result of these features, the expected amount of trait dissimilarity between recently diverged species increases with phylogenetic distance, but the signal of earlier divergence is ‘overwritten’ by recurrent evolution in the constrained trait space. A number of evolutionary processes may share the general properties of the OU model. The simplest interpretation of this model is of stabilizing selection around a single adaptive peak, where α is a measure of the strength of stabilizing selection (Hansen, 1997) and σ 2 describes the magnitude of departures from the optimum due to genetic drift. However, estimates of α from comparative datasets typically imply much stronger stabilizing selection (or much smaller effective population sizes) than is observed in most natural populations (Kingsolver et al., 2001; Harmon et al., 2010). Given the strong associations between morphology and trophic position that suggest adaptation to different diet types (Figure 2.3, Table 4.1), it seems unlikely that all rockfish are subject to stabilizing selection around a single peak representing an intermediate trophic position of around 3.7. The OU fit might also result from a random walk within a trait space with hard bounds, but the information available suggests that this is not the case for rockfish trophic position (see Results). While trophic position is certainly bounded, values estimated for rockfish do not approach the logical lower bound of 2.0 (for heterotrophs) or the practical upper bound of ∼5.5. The reality of rockfish trophic position evolution is assuredly more complex than any of these simple models. The random walk component of OU (and other) models is generally taken to reflect a large number of smaller events, such as change due to genetic drift, adaptive evolution where selective pressures vary over time, or response to selection on correlated traits. While the available data do not tell us which of these factors have been predominant, it is worth considering how and why trophic position might evolve.  4.4.3  Potential drivers of trophic position evolution  The phylogenetic analysis of the relationship between trophic position and morphology helps us to understand the functional morphology that evolves in concert with trophic position. However, these analyses tell us little about the factors that cause trophic position to change. These factors may include selective pressures due to biotic interactions or abiotic constraints, as well as nonadaptive processes such as genetic drift and rearrangements lower in the food web. Given the relationship between trophic position and size, TP may evolve passively due to any of the myriad potential evolutionary forces that influence organismal size (Peters, 1983). These may include selection for life history traits such as delayed maturation (Mangel et al., 2007) or selection for reduced vulnerability to gape-limited predators (Reimchen, 1991). Due to gape-limitation, any such change in body size will also change the maximum size of prey a 51  species can consume, potentially increasing or decreasing trophic position. However, much of the variation in trophic position among species is unrelated to body size, and gill raker morphology in particular is a stronger predictor of trophic position than size. If selection favours large body size but not higher trophic position, morphological adaptations can allow even a large-bodied species to efficiently feed on planktonic prey with low trophic positions. For example, the canary rockfish S. pinniger is the seventh largest species in my dataset (76 cm maximum total length), but has a relatively low trophic position (mean 3.40 based on isotopes) as well as the second most numerous gill rakers (mean 41.1). In contrast, the pinkrose rockfish S. simulator has a maximum TL of only 30 cm, but has both relatively few gill rakers (mean 29.5) and a relatively high trophic position (mean 3.97). Trophic position may also diverge as a result of behavioral differences between species in the absence of clear morphological change (Siemers et al., 2011). This partial decoupling of size and trophic position implies that trophic position does not evolve simply as a byproduct of selective pressures influencing body size. The trophic position of a consumer may instead evolve because the trophic positions of different potential diet items are correlated with features that affect their desirability as a resource. It may be favourable to consume lower trophic position resources because they have greater density or are easier to capture. Alternatively, selection may favour the consumption of organisms with higher trophic position if they are of higher quality, as measured using elemental stoichiometry (e.g. low C:P and C:N ratios) or content of important biomolecules such as lipids. This quantity-quality trade-off is most apparent in the contrast between plants and animals, but among autotrophs there is evidence that higher trophic position tends to correspond to higher quality (Fagan et al., 2002; Hendrixson et al., 2007). A potential explanation for the ubiquity of omnivory in food webs is that omnivorous species can get the best of both worlds by combining abundant but low quality resources with rarer, higher quality prey with higher trophic position (Denno and Fagan, 2003; Diehl, 2003). Interactions between species might contribute to trophic position evolution in several ways. Resource competition often drives divergence in diet between competing species (Schluter, 1994; Grant and Grant, 2006). In cases where the different diet items differ in their trophic position, competition may result in divergence along the axis of trophic position. Intraguild predation (combined competition and predation by one species on another; see Chapter 5) may result in trophic position divergence due to a combination of the quantity-quality trade-off and interspecific interactions. Intraguild predators may benefit from an ability to balance their intake of low-quality shared resources and high-quality intraguild prey (Diehl, 2003). The intraguild predator’s trophic position may also increase or decrease as a result of evolution by the intraguild prey (see Chapter 5). Rockfish are known to prey upon both smaller rockfish species and conspecific juveniles (Love et al., 2002), but the evolutionary impact of trophic interactions 52  within Sebastes is currently unknown.  4.4.4  Implications for the evolutionary recovery of food webs  Part of the motivation for investigating the evolution of trophic position is the potential extinction of many marine top predators due to overexploitation by humans (Pauly et al., 1998). If fishing pressures are ultimately relaxed at some point after the large-bodied top predators have become extinct, it would be useful to know how the evolutionary recovery or replacement of top predators might occur. To do this, we should understand the evolutionary dynamics of trophic position in trophically diverse groups of fish. Rockfish are not necessarily the most likely candidate to evolve top predators – their life history makes large rockfish in particular more vulnerable to extinction than many other fish – but the evolutionary analysis of rockfish trophic diversification may be a useful starting point to investigate how top predator recovery might occur. The simulations presented here (Figure 4.4) provide a thought experiment to begin to address this question of top predator evolution. I model the approach to a new adaptive peak of trophic position 5, starting with a species with a relatively high trophic position (4) and the maximum likelihood estimates of the σ 2 and α parameters of the OU model. The simulations imply a waiting time on the order of a few million years until a new adaptive peak would be reached, assuming no change in the magnitude of random fluctuations in the trait value (σ 2 ) and that the strength of selection to the new peak was comparable in magnitude to the value of the ‘rubber-band’ parameter α. The value of these simulations will depend on the factors that have caused the OU model to fit the rockfish dataset well. If the constraining force has been stabilizing selection due to competition with other species (e.g. top predators such as sharks), rockfish may indeed approach a new selective peak given the loss of these species. On the other hand, if the OU model fit results from biomechanical or genetic constraints intrinsic to rockfish, the group may not be capable of evolving further trophic diversity, and any selection for increased trophic position may not produce an evolutionary response. To more fully understand the past and potential future evolution of trophic position, it will be necessary to combine comparative analyses with population-level studies. By identifying the genetic variation underlying trophic position variation within species, and the potential for genetic correlations to aid or hinder a response to selection, we may be able to develop more robust predictions about the evolutionary potential for food web recovery.  53  Chapter 5  Intraguild predation drives evolutionary niche shift by threespine stickleback 5.1  Introduction  Food web interactions such as predation and resource competition are important agents of natural selection (Reznick and Endler, 1982; Schluter, 1994; Langerhans et al., 2004; Grant and Grant, 2006; Nosil and Crespi, 2006). The ensuing evolutionary responses of species modify their trophic interactions, and can consequently have cascading effects on food web structure and ecosystem function (Harmon et al., 2009; Bassar et al., 2010). The ecological consequences of one species evolution may alter selective pressures on it or other species, leading to further evolutionary change (Post and Palkovacs, 2009; Schoener, 2011). If these feedbacks are common and food webs are dynamically evolving networks, we need to understand how complex trophic interactions both cause and are altered by evolution. Intraguild predation is a widespread trophic interaction in which one species both feeds on a second species and competes with it for a shared resource (Polis et al., 1989; Polis and Holt, 1992; Arim and Marquet, 2004). Intraguild predation varies along a continuum between a tritrophic food chain and simple resource competition, depending on the relative reliance of the intraguild predator on the intraguild prey and the shared resource (Vandermeer, 2006). The double threat of competition and predation can have strong negative demographic effects on the intraguild prey (Polis et al., 1989; Diehl, 1995), which may be subject to strong natural selection. While the direction of these selective pressures are largely unknown, we can use the conditions under which theory predicts intraguild prey persistence (Holt and Polis, 1997; Vandermeer, 2006) to generate alternative hypotheses about evolutionary responses by intraguild prey species (Figure 5.1). Persistence of intraguild prey can be enhanced if the competitive and/or predatory impacts of 54  Trophic Position  Sculpin Sculpin Sculpin Stickleback  Benthic Invertebrates  Stickleback  Stickleback  Zooplankton  Increased Efficiency  Benthic Invertebrates  Zooplankton  Initial Sympatry  Benthic Invertebrates  Zooplankton  Niche Shift  Figure 5.1: Hypothesized evolutionary responses to intraguild predation and their ecological consequences. Vertical positions of species or trophic groups indicate their trophic position (a continuous measure of trophic level), and arrow widths indicate relative strength of feeding interactions. Following initial sympatry with sculpin, stickleback may become superior competitors for their shared benthic invertebrate resources (increased efficiency), likely leading to higher predation rates by sculpin on stickleback. Alternatively, stickleback may undergo a niche shift to rely more on the alternative resource (zooplankton), likely reducing predation and thus the trophic position of sculpin. the intraguild predator are weakened. For example, the intraguild prey is more likely to persist if it is the superior competitor for the shared resource (Holt and Polis, 1997; Morin, 1999; Vance-Chalcraft et al., 2007; Kondoh, 2008). A corresponding ‘increased efficiency’ hypothesis predicts that the intraguild prey will evolve traits that increase its ability to consume the shared resource. Alternatively, the intraguild prey may persist if it is subsidized by resources that are not shared with the intraguild predator (Daugherty et al., 2007; Holt and Huxel, 2007). A ‘niche shift’ that increases use of these alternative resources, and reduces resource overlap with the intraguild predator, is a second possible evolutionary response. Either increased efficiency on a shared resource or a niche shift to an alternative resource may occur under interspecific competition alone (Abrams, 1987), and should reduce the competitive impact of the intraguild predator on the intraguild prey. Additional evolutionary responses may reduce the predatory impact of the intraguild predator through inducible or constitutive defensive traits (Kratina et al., 2010), or through life history shifts (Reznick and Endler, 1982). Interestingly, a niche shift by the intraguild prey that includes a spatial habitat shift may simultaneously reduce competition and predation (Polis and Holt, 1992; Finke and Denno, 2006). Any evolutionary response by the intraguild prey has the potential to rearrange the food web. An interesting contrast between the ‘increased efficiency’ and ‘niche shift’ outcomes is the effective trophic level of the top predator (Figure 5.1). Under the increased efficiency model, the competitively inferior intraguild predator is itself more likely to persist if it increases its 55  consumption of the intraguild prey (Holt and Polis, 1997). In contrast, both a niche shift and the evolution of antipredator traits may increase the intraguild predator’s reliance on the shared resource. The intraguild predator’s trophic position is thus expected to increase if the intraguild prey evolves increased efficiency, and to decrease if it undergoes a niche shift. A niche shift may also cause the intraguild prey to derive its energy from another channel in the food web. These rearrangements of the food web have the potential to alter fundamental food web processes, for example by altering the symmetry of energy channels and the strength of top-down control (Pace et al., 1999; Rooney et al., 2008). Freshwater threespine stickleback (Gasterosteus aculeatus) of southwestern British Columbia occur with and without intraguild predators (Vamosi, 2003), making them an ideal system in which to test for evolutionary responses to intraguild predation. Stickleback repeatedly colonized small lakes from the ocean at the end of the last ice age (about 12,000 years ago). In contrast to a few lakes in which benthic and limnetic specialist species pairs have evolved, most lakes contain a single ‘solitary’ stickleback population that feeds on both benthic invertebrates in the littoral zone and zooplankton in the open water. We focus here on two types of solitary populations: ‘sympatric’ stickleback that occur in lakes with prickly sculpin (Cottus asper), a benthic intraguild predator, and ‘allopatric’ stickleback from lakes without sculpin. Sculpin are known to consume stickleback up to 60% of their body length (Moodie, 1972; Pressley, 1981). Sculpin also feed on one of the two major resources consumed by stickleback (benthic invertebrates), and have been found to reduce the foraging success and alter the resource use patterns of stickleback in field enclosures (Bolnick et al., 2010). Based on the scenarios we have outlined above, we predicted that stickleback sympatric with sculpin would either evolve increased efficiency on the shared benthic resource, or undergo a niche shift to rely more on non-benthic resources (e.g. zooplankton). We tested for either of these outcomes as well as antipredator morphological adaptations by comparing the phenotypes of wild stickleback from lakes with and without sculpin. We then carried out a mesocosm experiment to (i) compare the effects of sculpin addition on the fitness and resource use of representative sympatric and allopatric stickleback, and (ii) measure the food web consequences of stickleback character shifts. Both our comparative and experimental data support the hypothesis that stickleback have undergone a niche shift in the presence of their intraguild predator.  5.2 5.2.1  Materials and methods Comparative analysis of body shape  We analyzed geometric body shapes of stickleback from 24 populations in southwestern British Columbia to test for morphological changes in response to the intraguild predator (Table 5.1). We sampled five sympatric populations occurring with prickly sculpin, and ten allopatric populations 56  Table 5.1: Lake morphology and sample sizes of stickleback populations sampled for geometric body shape analysis Lake Priest (Van. I.)* Blackburn* Stowell* Dougan* Hoggan* Bullocks Kirk Klein Cranby Trout Paq North Brown Ambrose Cedar Enos Enos Paxton Paxton Priest Priest Little Campbell R. Oyster Lagoon Salmon R.  Population Type No sculpin No sculpin No sculpin No sculpin No sculpin No sculpin No sculpin No sculpin No sculpin No sculpin Sculpin Sculpin Sculpin Sculpin Sculpin Limnetic Benthic Limnetic Benthic Limnetic Benthic Marine Marine Marine  n 25 53 40 34 40 10 40 40 40 40 40 31 40 40 20 103 105 40 30 40 25 57 40 40  Elevation (m) 15 106 77 60 60 61 121 135 69 145 21 45 32 56 204 55 55 88 88 75 75 0 0 0  SA (ha) 2.3 3.7 5.6 10.0 19.7 9.4 8.3 13.5 44.6 7.6 12.1 12.8 18.8 29.8 31.0 17.6 17.6 17.0 17.0 44.3 44.3 -  Depth (m) 4.8 3.2 4.6 8.5 3.0 4.0 8.3 12.0 3.2 5.8 2.2 10.1 3.5 13.3 3.0 4.0 4.0 6.2 6.2 5.4 5.4 -  All lakes contain cutthroat trout (Oncorhyncus clarkii), and lakes marked with an asterisk (*) also contain introduced rainbow trout (O. mykiss). Lakes used as sources of allopatric (Trout Lake) and sympatric (Paq Lake) stickleback for the mesocosm experiment are in bold. The Little Campbell River population included 35 wild-caught and 22 pond-reared fish. SA = Surface Area. from lakes without sculpin. All lakes also contain cutthroat trout (Oncorhyncus clarkii), a predator of stickleback. As references to aid in the interpretion of any shape shift associated with sculpin presence, we also sampled benthic and limnetic stickleback from three lakes containing species pairs, as well three marine/anadromous populations representing the heavily armored ancestral form (Walker, 2000). The marine population from Little Campbell River is represented by two samples, one collected from the wild and the other raised in freshwater ponds at the University of British Columbia. Almost all populations occupy different watersheds and derive from separate colonizations of freshwater, so we treat them as statistically independent replicates. We sampled stickleback from each population for body shape analysis. All fish were 57  dorsal spines head 1 22  2  20 19 21  3  17  18 15 16  ectocoracoid  4  caudal peduncle 5  dorsal fin  14 13  12  pectoral fin  6  7  10  9  8  11 anal fin  Figure 5.2: Landmarks used for morphometric analysis of stickleback populations. euthanized with MS-222 (Argent Chemical Laboratories, Redmond WA, USA), preserved in 10% formalin for at least two weeks, stained with alizarin red to highlight bone, and stored in 37% isopropyl alcohol. We took digital photographs of the left side of each specimen and digitized 22 landmarks representing the positions of bony elements of the jaw, head and spines, and insertion points of the fins (Figure 5.2). These landmarks are similar to those used in previous shape analyses, are known to vary among stickleback populations, and in many cases have known functional significance (Walker, 1997, 2000; Spoljaric and Reimchen, 2007; Albert et al., 2008). Landmark positions on each photograph were digitized, centered and scaled to unit size using the programs tpsDig 1.40 and tpsRelw 1.44 (F.J. Rohlf; http://life.bio.sunysb.edu/morph/), and rotated to align with the average shape by minimizing the sum of squared distances over all homologous landmarks (Zelditch et al., 2004). These procedures resulted in 22 x and 22 y coordinates for each fish, adjusted for body size. As there were more x and y coordinates (44) than populations, it was necessary to reduce dimensionality of the data for analysis. We used a linear discriminant analysis (LDA), implemented in the MASS package (Venables and Ripley, 2002) in the R environment (R Development Core Team, 2009), to identify major axes of shape variation among populations relative to variation within populations. Importantly, for this analysis we did not identify specimens by population type, so the LDA did not bias our tests for differences between lakes with and without sculpin. We preferred LDA to alternatives such as principal components analysis (PCA) because LDA isolates those characters that vary among populations by finding axes that maximize variation among relative to within populations (Tabachnick and Fidell, 2001). In contrast, PCA can be heavily influenced by within-population variation, including measurement error and artifacts such as upward or downward bending of specimens. However, our results were qualitatively similar when we used PCA. We visualized vectors of shifts in individual landmarks (Figure 5.3) after first using PCA to remove specimen bending effects following Albert et al. (2008). 58  Our subsequent shape analyses used population means as replicates. Because of the relatively small number of populations (24), we restricted our multivariate tests to the first two discriminant functions, which explained 35 and 19 percent of the total variance among populations. These were the only shape axes that clearly grouped the populations according to type. We tested for differences among population types using multivariate and univariate ANOVAs on the population means of these two shape axes, focusing on the comparison of solitary populations from lakes with and without sculpin.  5.2.2  Mesocosm experiment  Our experiment crossed two treatments in a factorial design: sculpin addition (control vs. one sculpin added) and stickleback population of origin (sympatric with sculpin vs. allopatric). We established 36 experimental mesocosms in 1136 L plastic cattle tanks 1 m deep and 2 m wide. We randomly assigned eight tanks to each of our four primary treatments, and left the remaining four tanks free of all fish to test for overall fish effects. Tanks were filled with water and seeded with 10 L of benthic mud and filtrate from 80 L of water from nearby experimental ponds, containing many benthic and pelagic invertebrates and their propagules. We added 0.05 g KH2 PO4 and 1.0 g NaNO3 (equivalent to approximately 10 µg L−1 phosphorus and 160 µg L−1 nitrogen) to each cattle tank to stimulate primary production, then left the tanks for two weeks before adding fish. To provide stickleback with shade and refuge from predation, we suspended a 25 cm diameter open-ended cylinder of black 7 mm mesh (DuPont Vexar, Wilmington DE, USA) below the water surface. We collected 128 ‘allopatric’ stickleback from Trout Lake, 128 ‘sympatric’ stickleback from Paq Lake, and 16 sculpin from Sakinaw Lake, all in different watersheds on the Sechelt Peninsula. Trout and Paq Lakes are broadly similar in geographic location, productivity (summer surface chlorophyll a: 1-4 µg L−1 ; T. Ingram, unpublished data), surface area and depth (Table 5.1), and have stickleback with representative allopatric and sympatric body shapes (Figure 5.3). Fish were collected by minnow trap, transported to the University of British Columbia, and housed for two days in outdoor holding tanks. Eight stickleback were assigned to each mesocosm; each was anaesthetized with 0.1 g L−1 MS-222, then individually marked by subcutaneous injection of a two color combination of elastomer dye (Northwest Marine Technology, Shaw Island WA, USA). Stickleback were then weighed to 0.01 g, allowed to recover, and added to the mesocosms in early May, 2010. Sculpin were weighed and systematically assigned to the two sculpin addition treatments to ensure similar size distributions (allopatric: 13.90 ± 4.33 g; sympatric: 13.86 ± 3.64 g; mean ± SD). All sculpin were large enough (standard length range 9-16 cm) to ingest the majority of the stickleback (3-5.5 cm) in the tanks (Pressley, 1981). Sculpin were randomly introduced to tanks within their assigned 59  treatments two days after the stickleback. We surveyed tanks after two, four and six weeks to assess survivorship and growth of both species. We placed minnow traps in the tanks for two days, checking traps frequently and removing them at night to prevent predation by sculpin within the traps. Each recaptured fish was weighed again to measure its growth, and then returned it to its tank. To maintain approximately constant stickleback densities, we replaced fish that were not recaptured with additional fish from Paq and Trout Lakes. Because of differences in mortality (see Results), the number of replacement fish varied among treatments (allopatric control: 35; allopatric + sculpin: 77; sympatric control: 60; sympatric + sculpin: 62). Replacement fish were tagged and weighed, then distributed to return the densities in each tank to 8 fish (first round of replacements) and 7 fish (second round), though some densities were temporarily higher (up to 11) because fish were missed during a round of recaptures. Fish collected from the two lakes did not differ in size at initial stocking or the first round of replacements (range 0.88-0.92 g; p < 0.5), although in the second round allopatric stickleback were larger (allopatric: 0.76 ± 0.20 g; sympatric: 0.55 ± 0.28 g; mean ± SD; two-sample t-test: T101 = −4.49, p < 0.0001). This size difference might have introduced a bias toward increased predation rates on sympatric stickleback in tanks with sculpin, making our findings in the opposite direction conservative (see Results). During the third and final round of recaptures all stickleback and sculpin were removed from the tanks, weighed, euthanized with MS-222 and frozen. To ensure recovery of all fish, we trapped for three days then swept a large net through each tank several times. All 16 sculpin were recovered, though one had died a few days before the experiment ended; we calculated its growth rate based on its weight two weeks earlier. Stickleback reproduction occurred in many tanks, so we surveyed tanks for the presence of fry periodically during the experiment and five weeks after it ended. At this point, we collected approximately 25 stickleback fry from each of six cattle tanks without sculpin (three from each population) and transported them to the lab (see below).  5.2.3  Stickleback survival, growth and reproduction  As stickleback were added to the tanks at different times, we used two approaches to quantify survival. First, we performed a survival analysis using only the 256 stickleback present at the start of the experiment. We measured survival time as the number of days until a stickleback was last recaptured, and used the mean survival time in each tank as input for a parametric survival analysis with log-normal error structure, using the ‘surv’ and ‘psm’ functions in R (Crawley, 2007). We modeled the proportion of mesocosm populations predicted to become extinct over time as a function of sculpin addition, stickleback population of origin and their interaction, using likelihood ratio tests to assess significance of each term. As an alternative method allowing inclusion of all stickleback, we also categorized fish as ‘survivors’ if they were ever recaptured (i.e. survived at least one two-week census period), and ‘non-survivors’ otherwise. We quantified 60  the survival rate in each tank simply as the proportion of survivors. Unless otherwise noted, we transformed response variables when appropriate to meet the assumptions of regression, and modeled them as a function of sculpin addition, stickleback population of origin, and the sculpin × population interaction using a linear model in R. We calculated the growth rate of each stickleback and sculpin as its total mass gain divided by the number of days between introduction to the mesocosm and final recapture. We tested whether sculpin growth rates differed in tanks with sympatric vs. allopatric stickleback, predicting that greater consumption of stickleback would result in enhanced sculpin growth. For stickleback, we calculated the average growth rate within each tank, excluding 53 females that were visibly gravid at some point during the experiment. We tested whether sculpin affected the probability of successful reproduction in stickleback, and whether any such effect differed between populations. We used log-linear analyses to test whether sculpin addition, stickleback population or their interaction impacted the probability of successful reproduction.  5.2.4  Stickleback diet and impact on prey communities  We quantified diet by examining the stomach contents of all stickleback and sculpin recovered at the end of the experiment. We counted and identified diet items, and classified them as benthic (insect larvae, molluscs and stickleback eggs) or pelagic (zooplankton and chironomid pupae and adults that occur at the water surface). We used average lengths of diet items and published length-dry weight regression formulae (McCauley, 1984; Sample et al., 1993) to estimate the average mass of each item, and multiplied masses by prey counts to estimate the biomass of benthic and pelagic resources in each stomach. We then calculated the proportion of pelagic items by biomass in each stomach, and averaged this value within each tank. Four tanks with no stickleback alive at the end of the experiment were excluded from statistical analysis of diet. To test for differences in invertebrate biomass among treatments, we sampled zooplankton and benthic invertebrates directly from each mesocosm before the final removal of fish. We filtered 12 L of water from 10 cm below the surface through a 64 µm sieve, then stained and preserved zooplankton with 2-3 drops of acid Lugols solution. Zooplankton samples were identified to functional groups (usually genera) and counted under a dissecting microscope. Body lengths of up to 10 specimens per taxon per sample were measured with an ocular micrometer, and then used to estimate the biomass of zooplankton resources available to stickleback in each tank (McCauley, 1984). Occasional benthic taxa (chironomid larvae) present in zooplankton samples, and small taxa not consumed by stickleback (copepod nauplii and rotifers) were excluded from these biomass calculations. To sample benthic invertebrates, a small dipnet was used to collect two 120 cm2 scoops of benthic substrate (mud and decomposing leaves) from each tank. The substrate was rinsed 61  through a 500 µm sieve and searched for live organisms for 15 minutes. Invertebrates were stored in 95% ethanol for approximately 1 week, then counted and identified to functional groups. We measured benthic invertebrate biomass directly by drying samples for 24 h at 60 ◦ C and weighing them to 0.1 mg. We used the dry mass of the substrate sampled from each tank as a covariate to account for a strong relationship between substrate mass and benthic invertebrate biomass (F1,27 = 59.68, p < 0.001). We tested for overall top-down control of zooplankton and benthic invertebrate biomasses using a linear model contrasting the four fish-free mesocosms with the 32 mesocosms containing fish (this test was orthogonal to the contrasts in the main linear model, but was done separately to keep the main analyses consistent among response variables). We used two approaches to test for overall compositional changes in the mesocosm food web. First, we estimated the relative biomass of pelagic vs. benthic resources in the tank as the zooplankton biomass divided by the combined zooplankton and benthic invertebrate biomass. We estimated these values by extrapolating our sampled biomass based on the fraction of the tank sampled (∼12 of 1136 L for zooplankton and ∼240 of 31,400 cm2 for benthos), after first using a linear regression to adjust the benthic invertebrate biomasses to account for variation in the mass of substrate sampled. Second, we tested for compositional differences among treatments using redundancy analysis (Legendre and Legendre, 1998). We used the abundances of all invertebrate taxa (extrapolated as described above, and log transformed), including benthic and pelagic stickleback prey taxa as well as non-prey such as oligochaete worms, rotifers and copepod nauplii. We tested for significant effects of sculpin addition, stickleback type and their interaction using the permutation test (100 randomizations) in the R ‘vegan’ package.  5.2.5  Stickleback morphology  Finally, we characterized additional morphological traits of stickleback from Paq and Trout Lake to confirm that their phenotypes are representative of increased armor and pelagic foraging. We measured standard length and six ecomorphological traits from 30 wild-caught stickleback sampled from each lake at the same time as the experimental fish. We measured four defensive armor traits: the number of lateral plates on the left side of the body, and the lengths of the first and second dorsal spine and the left pelvic spine (Reimchen, 1994). We also recorded the number and maximum length of the gill rakers on the first gill arch, both of which are associated with increased plankton-feeding (Lavin and McPhail, 1986; Matthews et al., 2010). To confirm that any differences between populations are genetically based, we reared the ∼75 juvenile stickleback from each population born in the tanks without sculpin. These fish were reared in laboratory aquaria on mixed diets of brine shrimp nauplii, frozen Daphnia, and frozen bloodworms (i.e. both benthic and pelagic prey items). At an age of approximately 9 months, we sacrificed 20 fish from each population and measured the same traits as for the wild-caught fish. As the lab-reared fish were potentially cared for by their fathers for a period of time, this is not a 62  true common garden experiment. However, given the absence of predator cues early in life, and the low likelihood that juvenile fish foraged with adult stickleback, we feel that if divergent phenotypes were maintained in the lab it would provide strong evidence for genetically-based differences between populations. To test for differences between populations, we first computed population-specific slopes for the linear traits against standard length (pooling wild and lab-reared samples, as ANCOVAs revealed no differences in slope between groups; all p ≥ 0.25; Day et al., 1994). We calculated size-corrected values for all traits except gill raker number and lateral plate number, standardizing to a standard length of 40 mm. We used an ANOVA to test for effects of population, rearing environment (wild vs. lab-reared) and their interaction on these corrected trait values. Main effects of population would indicate overall differences, while population by rearing environment interactions would indicate that the magnitude of the difference changed due to a plastic response to the lab environment.  5.3 5.3.1  Results Comparative analysis of body shape  The linear discriminant analysis of shape coordinates revealed considerable shape variation among the five types of stickleback populations along the first two shape axes (Figure 5.3; multivariate ANOVA using population means as replicates: Wilks λ = 0.030, F4,20 = 22.9, p < 0.0001). The first axis clearly separated ancestral marine stickleback from derived freshwater forms, which are less streamlined with a larger head, larger eyes and a more posterior first dorsal spine. The second axis separated the benthic and limnetic species, representing the extreme body shapes found in freshwater. Limnetic-like stickleback have narrower bodies and larger eyes, jaws and dorsal and anal fins than benthic-like stickleback. Among solitary populations, stickleback sympatric with sculpin were consistently different in shape from allopatric stickleback (Figure 5.3; Wilks λ = 0.210, F1,13 = 22.6, p < 0.0001). This divergence was greatest on the second, benthic-limnetic axis (F1,13 = 48.9, p < 0.0001), but was also evident on the first, marine-freshwater axis (F1,13 = 5.21, p = 0.04). Compared with allopatric stickleback from lakes without sculpin, sympatric stickleback showed a conspicuous anterior shift in the first dorsal spine, a more slender body, more extensive dorsal and anal fins, and a larger ectocoracoid (the insertion point of the deep adductor muscle that powers pectoral fin swimming; Figure 5.3B). These shape differences are typically associated with increased predation and with sustained swimming and foraging in the pelagic habitat of lakes (Walker, 1997, 2000). These findings suggest that in the presence of an intraguild predator, stickleback have undergone morphological transitions associated with increased use of zooplankton, supporting the niche shift hypothesis. 63  A  limnetic  Shape Axis 2  4 sympatric (with sculpin)  ●  2  ●  marine  0  ● ●  Paq Lake ●  ●  allopatric (no sculpin) ●  ●  Pond  −2  ●  ●● ●●  Wild  ● ●  Trout Lake  benthic  −4 −6  −4  −2  0  Shape Axis 1  2  4  B  Figure 5.3: Body shape differences among the different types of stickleback populations. (A) Population means along the first two major shape axes from a linear discriminant analysis separating populations only. Symbols and convex hulls identify different types of stickleback populations. Arrows identify the two populations used in the mesocosm experiment, as well as wild-caught and pond-reared marine stickleback from the Little Campbell River marine population. (B) Differences in mean landmark positions between solitary populations in lakes without sculpin (base of arrow) and lakes with sculpin. Arrow lengths are multiplied by three for greater visibility.  64  5.3.2  Stickleback survival, growth and reproduction  We detected strong differences in the effect of sculpin addition on the fitness of sympatric and allopatric stickleback in the mesocosm experiment. Parametric survival analysis revealed a strong interaction whereby sculpin addition only reduced the survival of the allopatric stickleback population (sculpin × population interaction: F1,27 = 8.04, p = 0.005). Similarly, a lower proportion of allopatric stickleback survived between census periods in the presence (0.33 ± 0.08; mean ± SE) vs. absence (0.71 ± 0.07) of sculpin, whereas the survival rates of sympatric stickleback were similar in the presence (0.51 ± 0.04) or absence (0.48 ± 0.05) of sculpin (Figure 5.4A; sculpin × population interaction: p = 0.003). Sculpin growth was nearly threefold higher in mesocosms with allopatric stickleback (92.3 ± 14.0 mg day-1) than with sympatric stickleback (31.3 ± 7.8 mg day−1 ; p = 0.002), providing further evidence that sympatric stickleback experienced reduced predation by sculpin. While sculpin presence did not lead to any detectable increase in mortality of sympatric stickleback, it did cause a considerable reduction in their growth rate (Figure 5.4B). This population had lower growth rates overall (p = 0.002) and lower growth rates when sculpin were present (1.4 mg day−1 ± 0.6) than when sculpin were absent (7.9 mg day−1 ± 2.1; sculpin × population interaction: p = 0.032). In contrast, allopatric stickleback had comparably high growth rates in the control (9.9 ± 2.1 mg day−1 ) and sculpin addition treatments (11.2 ± 1.6 mg day−1 ). Sculpin addition also inhibited successful reproduction (nest-building, courtship, rearing and survival of fry) in both stickleback populations. Only 4 of 16 tanks with sculpin contained fry, compared to 13 of 16 in tanks without sculpin (log-linear analysis; d f = 2, G2 = 14.52, p = 0.0007; stickleback population and interactive effects were non-significant).  5.3.3  Stickleback diet and impact on prey communities  Diet of stickleback in the mesocosms, as revealed by their stomach contents at the end of the experiment, differed between populations (Figure 5.5). The proportion of pelagic prey (zooplankton) biomass in the diet was higher in the sympatric population (0.34 ± 0.043) than in the allopatric population (0.21 ± 0.033; p = 0.025), in agreement with the morphological evidence for a niche shift to the pelagic habitat. The main effect of sculpin (p = 0.24) and the sculpin × population interaction (p = 0.51) were non-significant, though there was a tendency for allopatric stickleback to have more zooplankton in their diets in tanks with sculpin present (Figure 5.5). Apart from one sculpin that had recently consumed one allopatric stickleback, sculpin stomachs contained greater than 99% benthic invertebrates by mass. The observed niche shift by stickleback toward increased zooplanktivory affected the biomass of their resources. Consistent with their more pelagic diet and morphology, sympatric stickleback reduced zooplankton biomass (6.4 ± 4.8 mg L−1 ) much more than did allopatric 65  ●  A  ●  Stickleback Survival  ●  0.75  ●  ●  ● ●  ●  0.5  ● ● ●  ● ● ●  ● ● ●  ●●  ●  ●  ● ● ●  ● ● ● ●  ● ●  0.25 ● ● ●  20  Stickleback Growth (mg day−1)  1  ●  B ●  ●  ●  15  ● ● ● ●  10  ● ●  ●  ●  ●  ● ●  ●  ●  ● ●  5  ● ●  ●  ●  ● ● ●  ●  ●  ● ● ● ●  0  0 Control  +Sculpin  Allopatric  Control  +Sculpin  ●  Control  Sympatric  +Sculpin  Allopatric  Control  +Sculpin  Sympatric  Figure 5.4: Survival and growth rates of sympatric and allopatric stickleback populations (from a lake with sculpin and a lake without sculpin, respectively), in experimental mesocosms with (+Sculpin) and without sculpin (Control). (A) Proportion of stickleback in each treatment surviving for at least one 14-day census period. (B) Average growth rates between the addition and final recapture of stickleback in each treatment (mean ± SE). stickleback (15.4 ± 7.6 mg L−1 ; p < 0.001; Figure 5.6A). In contrast, allopatric stickleback caused a slightly greater depletion of benthic invertebrate biomass, but only in tanks without sculpin (sculpin × population interaction: p = 0.08; Figure 5.6B). Sculpin addition led to an increase in zooplankton biomass (p = 0.025), but did not result in a detectable depletion of benthic invertebrate biomass (p = 0.39). Both zooplankton and benthic invertebrates were subject to top-down control in the mesocosms, as biomasses were higher in the four fish-free tanks than in tanks with fish (zooplankton biomass: F1,34 = 4.19, p = 0.048; benthic invertebrate biomass: F1,33 = 4.47, p = 0.042). Stickleback populations differed in their effect on the distribution of prey biomasses in the food web. By strongly suppressing zooplankton biomass and weakly enhancing benthic invertebrate biomass, sympatric stickleback reduced the pelagic proportion of the total invertebrate biomass in the mesocosm (0.0097 ± 0.004) relative to allopatric stickleback (0.036 ± 0.01; p < 0.0001). Sculpin addition increased the pelagic proportion of biomass (p = 0.043), while there was no sculpin × population interaction (p = 0.22). There was also no significant difference in the proportion of pelagic biomass in tanks with and without fish present (p = 0.31). We did not detect significant effects of stickleback population on the composition of invertebrate abundances using redundancy analysis (F1,28 = 0.98, p = 0.43), although sympatric stickleback did tend to negatively impact the densities of zooplankton taxa such as the herbivorous 66  Proportion Zooplankton in Diet  0.6  ● ● ● ● ● ● ●  0.4  ● ●  ●●  ●  ●  ● ● ●  ●  ●  ●  0.2  ● ●  ●  ●● ● ●  ● ●  ●  0  ●  Control  +Sculpin  Control  Allopatric  +Sculpin  Sympatric  Zooplankton Biomass (mg L−1)  ●  100  A  ● ●  ● ●  25  ● ●  ●  ● ●  ●  ●  ● ● ●● ● ● ●  5  ● ● ●  ●  ●  ●  ●  ● ● ●  ●  1  ●  ● ●  ● ●  ● ● ● ●  0.25 No Fish  Control  +Sculpin  Allopatric  Control  Benthic Invertebrate Biomass (mg cm−1)  Figure 5.5: Diets of stickleback recaptured at the end of the six-week mesocosm experiment. Values represent the estimated proportion of pelagic diet items (by biomass) in the stomachs, averaged within each mesocosm (mean ± SE).  +Sculpin  ● ●  50 ●  ● ●  25  ● ●  ● ● ●● ● ●  ● ● ●  10  ●  ● ● ● ●● ●  ● ● ● ●● ● ●  ● ●  ● ● ●  5 ● ●  2 No Fish  Sympatric  B  ●  Control  +Sculpin  Allopatric  Control  +Sculpin  Sympatric  Figure 5.6: Biomass of pelagic and benthic invertebrates in mesocosms at the end of the sixweek experiment. (A) Zooplankton biomass estimated from abundances, length measurements and length-dry weight regressions. (B) Dry mass of invertebrates collected from benthic substrate, corrected to remove a relationship with substrate mass (mean ± SE).  67  6  Allopatric Allopatric +Sculpin Sympatric Sympatric +Sculpin  ●  Acari ● ●  4 ●  RDA Axis 2  Cerio  2  Daph Chydo  Lepad  Cyclo  Diapha  K.quad  0 ChirPup  ●  ●  ●  ●  ChirLarv Chaob  K.coch Trich  ●  Ephem  Leca  −2  ● ● ●  Naup  Biv  ●  Oligo  Cala  ●  ●  ● ●  ●  −4 −4  −2  0  2  4  RDA Axis 1 Figure 5.7: Ordination plot from a redundancy analysis of the composition of the invertebrate community in the experimental mesocosms. The sculpin × population interaction was non-significant, and is not shown so as to highlight the main effects. Tanks are represented by symbols and treatments grouped by convex hulls. Labels indicate the vector from the origin (multiplied by eight for visibility) associated with each invertebrate taxon. cladoceran Daphnia sp. (Figure 5.7). Sculpin addition did significantly affect invertebrate composition (F1,28 = 2.60, p = 0.01), in particular by increasing the abundance of cyclopoid copepod copepodites and nauplii, while decreasing the abundance of rotifers. There was no sculpin × population interaction (F1,28 = 1.14, p = 0.29), and this effect was dropped from the analysis to highlight the main effects (Figure 5.7).  5.3.4  Stickleback morphology  Sympatric stickleback from Paq Lake have longer dorsal and pelvic spines, more lateral plates and longer gill rakers than allopatric stickleback from Trout Lake (Figure 5.8), consistent with 68  higher predation risk and a more pelagic habitat and diet (Lavin and McPhail, 1986; Reimchen, 1994; Matthews et al., 2010). These differences appear to be genetically based, as the fish reared in the lab showed as much phenotypic differentiation between populations as the wild-caught fish. We did not detect any interactive effects of population and rearing environment whereby a trait that differed between populations in the wild was significantly less different in the lab-reared fish. There were rearing environment effects on numbers of gill rakers and lateral plates; measurements were taken by different individuals close to a year apart, so it is possible that these effects are due to measurement bias. However, any bias should be consistent between populations, making population and population by rearing environment interaction effects valid. The lack of interactive effects for most traits indicates that the morphological differences between populations have a largely genetic basis.  5.4  Discussion  We have presented several lines of evidence that stickleback have evolved in response to an intraguild predator, prickly sculpin. The morphological comparison suggested that the presence of the intraguild predator leads to changes in stickleback body shape associated with enhanced defenses and a shift to a more pelagic habitat. The mesocosm experiment indicated that a sympatric population that evolved in the presence of sculpin is less vulnerable to predation and more planktivorous than an allopatric population that evolved in the absence of sculpin. Overall, our results show that the response of stickleback conforms to the niche shift rather than the increased efficiency hypothesis, and suggest that this niche shift impacts the structure of the food web. These conclusions come with three important caveats. First, the morphological results are based on observational data, meaning we cannot rule out other differences between lakes with and without sculpin that may contribute to the shift in stickleback morphology. Second, mesocosm experiments need to be carried out with additional populations to confirm that the ecological differences we detected are consistently associated with the presence or absence of sculpin in lakes. The general similarity of the lakes used in the experiment and the repeatability of the shape shift in the presence of sculpin increase our confidence that sculpin presence is the primary driver of the strong differences between populations. The third caveat to consider is the potential role of phenotypic plasticity in the ecological differences between wild-caught sympatric and allopatric stickleback. The morphological differences between our focal populations persist in the lab, implying that they have a genetic basis (Figure 5.8). We also expect the shape differences between populations to be largely under genetic control, because both rearing experiments (Figure 5.3; Spoljaric and Reimchen, 2007) and genetic mapping studies (Schluter et al., 2004; Albert et al., 2008) indicate that shape differences among stickleback populations are heritable and affected by many genes. There is 69  0.95 0.90 0.85  ●  ● ●  Lab  ●  ●  ● ●  ●  Wild  Lab  Sympatric  19  4 Wild  Allopatric  21  22  ●  Pop: p = 0.47 Env: p = 0.0001 PxE: p = 0.66  20  6  ●  Gill Raker Number  7  ●  Pop: p < 0.0001 Env: p = 0.002 ● PxE: p = 0.10  5  4.2 3.6 3.4 3.2  ●  Lateral Plate Number  ● ●  3.8  4.0  Pop: p < 0.0001 Env: p = 0.52 PxE: p = 0.18  3.0  Second Dorsal Spine (mm)  ●  0.70  ●  Pop: p < 0.0001 Env: p = 0.72 PxE: p = 0.69  0.80  5.0 ●  0.75  5.5  ●  Gill Raker Length (mm)  2.6  ●  ● ●  4.5  2.8  3.0  ●  4.0  Pelvic Spine (mm)  3.4 3.2  ●  Pop: p < 0.0001 Env: p = 0.72 PxE: p = 0.39  2.4  First Dorsal Spine (mm)  Pop: p < 0.0001 Env: p = 0.53 PxE: p = 0.07  Wild  Lab  Allopatric  Wild  Lab  Sympatric  Wild  Lab  Allopatric  Wild  Lab  Sympatric  Figure 5.8: Morphology of wild-caught and lab-reared fish from the allopatric (Trout Lake) and sympatric (Paq Lake) populations used in the experiment (mean ± SE). Linear measurements (spine and gill raker lengths) are size-corrected to a standard length of 40 mm. N = 30 wild-caught and 20 lab-reared fish per lake. Significance levels from ANOVAs are given for population effects (Pop), rearing environment effects (Env), and their interaction (PxE). little evidence of short-term behavioral plasticity in our results, as allopatric stickleback did not avoid predation when sculpin were added, and sympatric stickleback did not alter their diet in the absence of sculpin. However, both genetic differences and early life exposure to predators appear to contribute to the adult behavior and morphology of stickleback populations experiencing divergent predation regimes (Dingemanse et al., 2009). Studies using lab-reared fish are therefore needed to estimate the contribution of genetic divergence and plasticity to the ecological differences between populations from lakes with and without sculpin. Our data suggest that evolution of an intraguild prey can reduce the predatory impact of its intraguild predator. While the mechanisms remain to be tested, enhanced morphological defenses such as spines and lateral plates can reduce the vulnerability of stickleback to predation by larger  70  fish (Reimchen, 1994). In addition, sympatric stickleback may have enhanced predator inspection and avoidance behaviors (Dingemanse et al., 2009), and the niche shift to the open water habitat may itself reduce encounters between intraguild predator and prey (Finke and Denno, 2006). Sympatric stickleback experienced reduced growth rates when sculpin were added, indicating that non-predatory effects of sculpin increased when their predatory effect decreased. These effects may result from a behavioral response by stickleback to the presence of sculpin. For example, a trade-off between foraging and predation risk may result in sympatric stickleback feeding less frequently or on less profitable resources in the presence of sculpin (Milinski and Heller, 1978; Werner and Hall, 1988). Alternatively, sympatric stickleback may experience stronger resource competition from sculpin, which is counterintuitive given that they feed more on pelagic resources than allopatric stickleback. However, if sculpin consume more benthic invertebrates because they are unable to feed on sympatric stickleback, they may be stronger competitors for this resource, which still comprised 66% of the diet of sympatric stickleback (Figure 5.5). In either case, stickleback may be largely excluded from benthic habitats in lakes with sculpin. Our results also shed some light on the factors promoting speciation in threespine stickleback. The evolution of benthic and limnetic species pairs has been driven in part by resource competition between stickleback (Schluter, 1994; Pritchard and Schluter, 2001), and may have been facilitated by predation from cutthroat trout (Vamosi and Schluter, 2002; Rundle et al., 2003). In contrast to these individual effects of competition and predation, intraguild predation appears to inhibit speciation, as species pairs have not evolved in any lakes containing sculpin (Vamosi, 2003). Predation and/or competition from sculpin may reduce the profitability of the benthic niche, limiting the potential for persistent divergent selection. The effects of sculpin on stickleback reproduction may also constrain diversification by changing male stickleback nesting behavior (Pressley, 1981), which is thought to be an important component of assortative mating in lakes with species pairs (McPhail, 1994). Further investigations in stickleback and other species may reveal whether different types of trophic interactions (e.g. intraguild predation vs. predation) generally have different effects on speciation. We have focused on the effects of sculpin on stickleback rather than the reverse, but intraguild predators may also evolve in response to the presence or evolution of their intraguild prey. Our data suggest that stickleback evolution in response to intraguild predation has led to a reduction in the trophic position of sculpin. Stickleback typically have trophic positions of 3.4-4 (Matthews et al., 2010), whereas many benthic invertebrates are herbivorous (i.e. trophic position 2), so sculpin that no longer feed on stickleback may be at least a full trophic level closer to the base of the food web. The threefold lower growth of sculpin in tanks with sympatric stickleback compared with allopatric stickleback suggests that this dietary change has negative fitness consequences for the sculpin. One possible consequence is that natural selection would favor 71  sculpin that are more effective piscivores. Experiments measuring selection on co-occurring sculpin and stickleback populations will be necessary to more fully characterize the evolutionary dynamics of this interaction. Our results also contribute to a growing body of studies demonstrating that adaptive evolution can have consequences for the structure of experimental food webs (Harmon et al., 2009; Palkovacs and Post, 2009; Bassar et al., 2010) that may scale up to affect whole ecosystems (?). By depleting zooplankton, sympatric stickleback shifted the balance of biomass toward a more benthic-dominated food web (although sculpin presence partially compensated for this effect). The lack of strong effects on benthic resources may result from high spatial variability in benthic invertebrate biomass within tanks, or interference between the intraguild predator and prey may have kept the effect of fish on benthic invertebrates relatively constant across treatments. We believe that the effect of stickleback population on zooplankton biomass is the result of differential foraging, for two reasons. First, sympatric stickleback contained more zooplankton biomass in their stomachs despite the low zooplankton availability, indicating that they preferentially fed on zooplankton. Second, while treatments did differ in stickleback biomass at some stages of the experiment, there were consistently more zooplankton in tanks with allopatric stickleback whether there were higher stickleback densities (in tanks without sculpin) or lower stickleback densities (in tanks with sculpin) than in either treatment with sympatric stickleback. Thus, the effect on the food web does appear to result from diet differences between stickleback populations. In addition to the effect on zooplankton biomass, stickleback evolution may alter several aspects of trophic structure. By reducing predation by sculpin, evolution by stickleback appears to have shortened benthic food chains by removing or weakening an intermediate trophic link (Post and Takimoto, 2007). In contrast, the niche shift to more pelagic habitats may increase encounters with predatory cutthroat trout, possibly lengthening pelagic food chains and contributing to the evolution of increased armor (Vamosi and Schluter, 2002). As stickleback become more planktivorous, benthic and pelagic channels of the lake food web may also become partially decoupled (Figure 5.1). Food chain length and energy channel coupling are key determinants of food web dynamics and stability (Pace et al., 1999; Rooney et al., 2008), suggesting that an evolutionary response to intraguild predation may have broad ecological consequences. We have provided some of the first empirical evidence for the direction of evolution in response to intraguild predation. Niche shifts, instead of increased efficiency, may be a common evolutionary response of intraguild prey, especially when alternative resources are spatially segregated from the intraguild predator. By altering the competitive and predatory components of the interaction, evolutionary responses to intraguild predation can result in a dynamic restructuring of natural food webs. 72  Chapter 6  Conclusion In the preceding chapters, I have addressed questions about trophic niche evolution using comparative, experimental and theoretical approaches. Given the breadth of topics my research has covered, some of my key results have diverged somewhat from the questions related to trophic position evolution raised in the introduction. In Chapter 2, I discover a relationship between speciation and depth divergence in rockfish using phylogenetic comparative methods. In Chapter 3, the evolutionary food web assembly model produces some intriguing emergent relationships between the strength of foraging trade-offs and the structure and temporal stability of the food web. Finally, in Chapter 5, my collaborators and I document an evolutionary response by stickleback to intraguild predation that has consequences for the rest of the food web. These results are explored at length in the the relevant chapters. In this section, I discuss some of the common threads that connect these chapters, and attempt to synthesize the current state of knowledge about trophic niche evolution and outline some goals for future research.  6.1  Is trophic niche evolution involved in speciation?  My first question was whether the evolution of trophic position in rockfish is generally associated with speciation events. There are a number of reasons this might be the case: for example, speciation may involve disruptive selection on diet, intraguild predation between species, or adaptation to new prey types following habitat divergence. My comparative analysis of rockfish using my ψ parameter suggest that rockfish speciation is not generally associated with changes in trophic position. While there was considerable uncertainty around my estimates of ψ, the most likely parameter value indicated that speciational change did not contribute to the evolution of trophic position or the morphological traits it is most associated with: body size and gill raker number and length. If not rockfish, are there other systems in which trophic position evolution is more likely to be involved in speciation? When ecological speciation with gene flow occurs, habitat divergence 73  may be more common than divergence in local resource use. This is partly because adaptation along the gradient can intensify disruptive selection due to competition between spatially and phenotypically proximate individuals, and partly because spatial habitat separation contributes to the reduction of gene flow (Doebeli and Dieckmann, 2003). If speciation does involve divergence in diet within a habitat, this diet divergence may still be ‘horizontal’ and not lead to divergence in trophic position. Alternatively, trophic position may diverge during speciation, but as a byproduct of horizontal diet divergence between spatially separated food web ‘channels’. For example, in stickleback species pairs, limnetics consistently have higher trophic positions than benthics, due to the longer food chains in the pelagic channel of lake food webs (Matthews et al., 2010). One situation in which the vertical aspect of the niche might mechanistically be involved in speciation is if predation between conspecifics is one of the drivers of divergence. Intraguild predation from another species seems to constrain speciation in stickleback (Chapter 5), but intraguild predation by conspecifics (i.e. cannibalism) has the potential to promote divergence. In size-structured populations, larger individuals can prey on smaller individuals, but may be inferior competitors for shared resources due to the allometric scaling of foraging and metabolic rates (Claessen et al., 2000). Thus, disruptive selection may favour both larger and smaller body sizes, and favour the evolution of two morphs within the population. In a system with suitable reproductive traits (e.g. where mate choice is based on body size; Boughman et al., 2005) and ecological conditions (Holt and Polis, 1997), intraguild predation may even favour the evolution of reproductive isolation and the subsequent coexistence of two species.  6.2  How does trophic position evolve in the course of adaptive radiation?  The evolutionary food web assembly model described in Chapter 3 led to a number of insights about the potential relationships between trophic structure, species turnover, and macroevolutionary patterns of trait evolution. These relationships arise as a result of variation in the strength of the foraging trade-offs assumed. The strength of trade-offs may vary as a function of a feature of the environment, such as the degree of structural complexity. Trade-off strength may also vary among groups of organisms as a result of genetic or phenotypic characteristics such as the degree of modularity in trophic morphology or the extent to which body size determines trophic position. Clades may therefore vary in their propensity to evolve new trophic levels. One of several predictions of this model is that in clades that occupy only a few discrete and stable trophic niches, macroevolutionary patterns of trophic position should fit an ‘early burst’ model. In contrast, clades that are highly omnivorous are more likely to fit either a random walk or a model of recurrent evolution in a constrained trait space. The latter model provided the best fit to my rockfish trophic position and trophic morphology data (Chapter 4). This is consistent with what is known of rockfish diets, as many species are generalists that consume a variety of 74  prey of different trophic position. A proper test of the predicted relationship between degree of omnivory and macroevolutionary pattern will require a comparison of clades such as rockfish to clades subject to stronger trade-offs that largely occupy discrete trophic levels.  6.3  How does evolution modify food web interactions?  Chapter 5 presents evidence for repeated evolutionary divergence in threespine stickleback populations associated with the presence of an intraguild predator, prickly sculpin. The trajectory of morphological change suggests that both competition for benthic prey and predation have contributed to this divergence. Our mesocosm experiment confirms that a population sympatric with sculpin has both reduced vulnerability to predation and lower diet overlap with sculpin than a population that does not share an evolutionary history with sculpin. While we have not yet made quantitative measures of species’ trophic positions in this system, we can draw some provisional conclusions about how the evolutionary response to sculpin impacts the trophic structure of the food web. The trophic position of sculpin is likely to decline – perhaps as much as a full trophic level – as adaptation by stickleback makes them essentially unavailable as prey. A subtler effect may be an increase in stickleback trophic position with greater dependence on zooplankton and thus membership in longer pelagic food chains (Matthews et al., 2010). These predicted shifts in trophic position resulting from stickleback evolution should be tested for directly in this system. Stable isotope and diet analyses in wild populations would provide a partial answer, and longer term experiments tracking the evolution of stickleback following the invasion of sculpin would provide a much more complete treatment of this coevolutionary process. Evolutionary shifts resulting from species interactions are likely to alter trophic structure in many other systems. Freshwater fish provide the best-studied systems in which evolution has been shown to feed back onto food web processes. For example, the presence of predators drives evolutionary change in guppy (Poecilia reticulata) morphology and life history (Reznick and Endler, 1982; Reznick et al., 1990). Concurrent with these changes, high-predation guppies have shifted to feeding more on benthic invertebrates and less on algae, changing their impact on food web and ecosystem processes (Palkovacs et al., 2009; Bassar et al., 2010). In alewifes (Alosa harringtonensis) in the eastern United States, the landlocking of some populations has led to depletion of large zooplankton and an eco-evolutionary feedback that has substantially altered community structure (Post and Palkovacs, 2009; Palkovacs and Post, 2009). The evolution of trophic position in each of these focal populations is likely to contribute to their impacts. Ultimately, these systems – and eventually similar studies in marine and terrestrial study systems – should help us to understand when and how trophic niche evolution matters for ecosystem processes.  75  6.4  Future directions in the study of trophic niche evolution  I envision two major directions for future research into the topics I have discussed in my thesis. First, a comparative approach to the evolution of trophic position should be extended to many more clades to identify how patterns vary as a function of both organismal and environmental variables. Second, systems should be developed in which to investigate experimentally the evolution of trophic position under different conditions. Many clades span one or more trophic levels, allowing both analysis of rates and patterns of trophic position evolution and evaluation of the factors involved in producing these patterns. With the increasing availability of molecular phylogenies, it should be possible to assemble many comparative trophic position datasets using either stable isotopes or diet analyses. These datasets can then be subjected to phylogenetic analyses, both the simple model fitting used in Chapter 4 and eventually more complex models that explicitly incorporate the processes we are most interested in. An expanded comparative analysis of trophic position evolution should allow us to address a number of important questions about the evolution of trophic niches and food web structure. We can test whether taxonomic groups differ in their propensity to evolve vertical trophic diversity. We can also examine whether the rate of trophic position evolution varies with the average trophic position of a clade. Transitions between the first and second trophic levels are extremely rare (carnivorous plants being a partial exception), but even among animals it may be more difficult to evolve between trophic position 2 (herbivore) and 3 (primary carnivore) than between trophic position 3 and 4. This might be expected if the transition between feeding on plants and on animals is more difficult than between feeding on animals at different trophic levels. Finally, we can investigate whether the rate and pattern of trophic position evolution varies with features of the environment, such as the degree of structural complexity, or of organisms, such as their degree of omnivory. Such a comparative approach will allow us to identify factors that vary among clades and are correlated with the rate and extent of vertical diversification. An experimental approach to trophic niche evolution is an important step toward testing causal relationships between these factors and the rate of trophic position evolution. Two main experimental approaches have been used to investigate the evolution of trophic interactions to date. First, experimental introductions have demonstrated evolutionary responses to trophic interactions in a few generations in natural systems (Reznick et al., 1990; Losos et al., 2004). Second, microcosm experiments have allowed the demonstration that rapid evolution can alter predator-prey interactions (Yoshida et al., 2003; Becks et al., 2010). A next step is to extend these studies of predator-prey interactions to experimentally investigate the evolution of trophic position itself. As I have argued previously in this thesis, intraguild predators and intraguild prey will be promising target organisms in which to examine rapid evolution of food web interactions and 76  trophic position. While model systems for studying evolution in intraguild predation interactions have yet to be developed, a considerable amount of work has investigated the ecology of intraguild predation in microcosm populations. Much of this work has been carried out in ciliated protozoans (Holyoak and Sachdev, 1998; Morin, 1999), whose short generation time should make them suitable for experimental evolution studies as well. The development of experimental model intraguild predation systems will allow us to test a number of important evolutionary hypotheses. We can test whether evolution by intraguild prey tends to produce ecological conditions that favour stability, such as increased competitive efficiency or niche shifts (Chapter 5), and whether coevolution by the intraguild predator increases or decreases any such stabilizing effect. It should also be possible to carry out direct tests of mechanisms that might promote omnivory and trophic niche evolution, such as a trade-off between resource quality and availability. Finally, we may be able to test whether intraguild predation within a population can be a driver of disruptive selection and ultimately speciation. Such a discovery would provide a direct link between the buildup of species diversity and the evolution of trophic structure. It will be some time before we develop a complete conceptual model of the evolutionary dynamics of organisms within food webs, but these and other studies will move us along the path toward this important goal.  77  Bibliography Abrams, P. A., 1987. Alternative models of character displacement and niche shift. I. Adaptive shifts in resource use when there is competition for nutritionally nonsubstitutable resources. Evolution 41:651–661. Ackerly, D. D., 2009. Conservatism and diversification of plant functional traits: Evolutionary rates versus phylogenetic signal. Proc Natl Acad Sci USA 106:19699–19706. Ackerly, D. D. and W. K. Cornwell, 2007. A trait-based approach to community assembly: partitioning of species trait values into within- and among-community components. Ecol Lett 10:135–145. Ackerly, D. D., D. W. Schwilk, and C. O. Webb, 2006. Niche evolution and adaptive radiation: testing the order of trait divergence. Ecology 87:S50–S61. Ackermann, M. and M. Doebeli, 2004. Evolution of niche width and adaptive diversification. Evolution 58:2599–2612. Albert, A. Y. K., S. Sawaya, T. Vines, A. K. Knecht, C. T. Miller, B. R. Summers, S. Balabhadra, D. M. Kingsley, and D. Schluter, 2008. The genetics of adaptive shape shift in stickleback: pleiotropy and effect size. Evolution 62:76–85. Anderson, D. R., K. P. Burnham, and W. L. Thompson, 2000. Null hypothesis testing: problems, prevalence, and an alternative. J Wildl Manage 64:912–923. Arim, M. and P. A. Marquet, 2004. Intraguild predation: a widespread interaction related to species biology. Ecol Lett 7:557–564. Barluenga, M., K. N. St¨olting, W. Salzburger, M. Muschick, and A. Meyer, 2006. Sympatric speciation in Nicaraguan crater lake cichlid fish. Nature 439:719–723. Barnes, C., D. M. Bethea, R. D. Brodeur, J. Spitz, V. Ridoux, C. Pusinari, B. C. Chase, M. E. Hunsicker, F. Juanes, A. Kellerman, J. Lancaster, F. Menard, F. X. Bard, P. Munk, J. K. Pinnegar, F. S. Scharf, R. A. Rountree, K. I. Stergiou, C. Sassa, A. Sabates, and S. Jennings, 2008. Predator and prey body sizes in marine food webs. Ecology 89:881. Barraclough, T. G. and A. P. Vogler, 2000. Detecting the geographical pattern of speciation from species-level phylogenies. Am Nat 155:419–434. Bascompte, J., C. J. Meli´an, and E. Sala, 2005. Interaction strength combinations and the overfishing of a marine food web. Proc Natl Acad Sci USA 102:5443–5447. 78  Bassar, R. D., M. C. Marshall, A. L´opez-Sepulcre, E. Zandon`a, S. K. Auer, J. Travis, C. M. Pringle, A. S. Flecker, S. A. Thomas, D. F. Fraser, and D. N. Reznick, 2010. Local adaptation in Trinidadian guppies alters ecosystem processes. Proc Natl Acad Sci USA 107:3616–3621. Beckerman, A. P., O. L. Petchey, and P. H. Warren, 2006. Foraging biology predicts food web complexity. Proc Natl Acad Sci USA 103:13745–13749. Becks, L., S. P. Ellner, L. E. Jones, and N. G. Hairston, 2010. Reduction of adaptive genetic diversity radically alters eco-evolutionary community dynamics. Ecol Lett 13:989–997. Blomberg, S. P., T. Garland Jr, and A. R. Ives, 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution 57:717–745. Bokma, F., 2008. Detection of “punctuated equilibrium” by Bayesian estimation of speciation and extinction rates, ancestral character states, and rates of anagenetic and cladogenetic evolution on a molecular phylogeny. Evolution 62:2718–2726. Bolnick, D. I., T. Ingram, W. E. Stutz, L. K. Snowberg, O. L. Lau, and J. S. Paull, 2010. Ecological release from interspecific competition leads to decoupled changes in population and individual niche width. Proc R Soc Lond B 277:1789–1797. Boughman, J. W., H. D. Rundle, and D. Schluter, 2005. Parallel evolution of sexual isolation in sticklebacks. Evolution 59:361–373. Bowers, M. A. and J. H. Brown, 1982. Body size and coexistence in desert rodents: chance or community structure? Ecology 63:391–400. Branch, T. A., R. Watson, E. A. Fulton, S. Jennings, C. R. Mcgilliard, G. T. Pablico, D. Ricard, and S. R. Tracey, 2010. The trophic fingerprint of marine fisheries. Nature 468:431–435. Brose, U., L. Cushing, E. L. Berlow, T. Jonsson, C. Banaˇsek-Richter, L.-F. Bersier, J. L. Blanchard, T. Brey, S. R. Carpenter, M.-F. Cattin Blandenier, J. E. Cohen, H. A. Dawah, T. Dell, F. Edwards, S. Harper-Smith, U. Jacob, R. A. Knapp, M. E. Ledger, J. Memmott, K. Mintenbeck, J. K. Pinnegar, B. C. Rall, T. Rayner, L. Ruess, W. Ulrich, P. Warren, R. J. Williams, G. Woodward, P. Yodzis, and N. D. Martinez, 2005. Body sizes of consumers and their resources. Ecology 86:2545. Brose, U., T. Jonsson, E. L. Berlow, P. Warren, C. Banasek-Richter, L.-F. Bersier, J. L. Blanchard, T. Brey, S. R. Carpenter, M.-F. C. Blandenier, L. Cushing, H. A. Dawah, T. Dell, F. Edwards, S. Harper-Smith, U. Jacob, M. E. Ledger, N. D. Martinez, J. Memmott, K. Mintenbeck, J. K. Pinnegar, B. C. Rall, T. S. Rayner, D. C. Reuman, L. Ruess, W. Ulrich, R. J. Williams, G. Woodward, and J. E. Cohen, 2006a. Consumer–resource body-size relationships in natural food webs. Ecology 87:2411–2417. Brose, U., R. J. Williams, and N. D. Martinez, 2006b. Allometric scaling enhances stability in complex food webs. Ecol Lett 9:1228–1236. Burford, M. O., 2009. Demographic history, geographical distribution and reproductive isolation of distinct lineages of blue rockfish (Sebastes mystinus), a marine fish with a high dispersal potential. J Evol Biol 22:1471–1486. 79  Caldarelli, G., P. G. Higgs, and A. J. McKane, 1998. Modelling coevolution in multispecies communities. J Theor Biol 193:345–358. Cardinale, B. J., D. S. Srivastava, J. E. Duffy, J. P. Wright, A. L. Downing, M. Sankaran, and C. Jouseau, 2006. Effects of biodiversity on the functioning of trophic groups and ecosystems. Nature 443:989–992. Carroll, S. P., A. P. Hendry, D. N. Reznick, and C. W. Fox, 2007. Evolution on ecological time-scales. Funct Ecol 21:387–393. Case, T. J., 1990. Invasion resistance arises in strongly interacting species-rich model competition communities. in Proc Natl Acad Sci USA, Pp. 9610–9614. Cattin, M.-F., L.-F. Bersier, C. Banasek-Richter, R. Baltensperger, and J.-P. Gabriel, 2004. Phylogenetic constraints and adaptation explain food-web structure. Nature 427:835–839. Claessen, D., de Roos A M, and L. Persson, 2000. Dwarfs and giants: cannibalism and competition in size-structured populations. Am Nat 155:219–237. Cohen, J. E. and C. M. Newman, 1985. A stochastic theory of community food webs: I. models and aggregated data. Proc R Soc Lond B 224:421–448. Crawley, M. J., 2007. The R book. John Wiley and Sons, Chichester, UK. Darimont, C. T., S. M. Carlson, M. T. Kinnison, P. C. Paquet, T. E. Reimchen, and C. C. Wilmers, 2009. Human predators outpace other agents of trait change in the wild. Proc Natl Acad Sci USA 106:952–954. Daugherty, M. P., J. P. Harmon, and C. J. Briggs, 2007. Trophic supplements to intraguild predation. Oikos 116:662–677. Day, T., J. Pritchard, and D. Schluter, 1994. Ecology and genetics of phenotypic plasticity: a comparison of two sticklebacks. Evolution 48:1723–1734. Denno, R. F. and W. F. Fagan, 2003. Might nitrogen limitation promote omnivory among carnivorous arthropods? Ecology 84:2522–2531. Diamond, J. M., 1986. Evolution of ecological segregation in the New Guinea montane avifauna. Pp. 98–125, in J. M. Diamond and T. J. Case, eds. Community Ecology. Harper & Row, Cambridge, MA. Dieckmann, U. and M. Doebeli, 1999. On the origin of species by sympatric speciation. Nature 400:354–357. Diehl, S., 1995. Direct and indirect effects of omnivory in a littoral lake community. Ecology 76:1727–1740. ———, 2003. The evolution and maintenance of omnivory: dynamic constraints and the role of food quality. Ecology 84:2557–2567.  80  Dingemanse, N. J., F. Van der Plas, J. Wright, D. R´eale, M. Schrama, D. A. Roff, E. Van der Zee, and I. Barber, 2009. Individual experience and evolutionary history of predation affect expression of heritable variation in fish personality and morphology. Proc R Soc Lond B 276:1285–1293. Doebeli, M. and U. Dieckmann, 2003. Speciation along environmental gradients. Nature 421:259–264. Drenner, R. W., J. R. Mummert, F. deNoyelles Jr, and D. Kettle, 1984. Selective particle ingestion by a filter-feeding fish and its impact on phytoplankton community structure. Limnol Oceanogr 29:941–948. Drossel, B., P. G. Higgs, and A. J. McKane, 2001. The influence of predator–prey population dynamics on the long-term evolution of food web structure. J Theor Biol 208:91–107. Drossel, B., A. J. McKane, and C. Quince, 2004. The impact of nonlinear functional responses on the long-term evolution of food web structure. J Theor Biol 229:539–548. Drummond, A. J. and A. Rambaut, 2007. BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol 7:214. Duffy, J. E., B. J. Cardinale, K. E. France, P. B. McIntyre, E. Th´ebault, and M. Loreau, 2007. The functional role of biodiversity in ecosystems: incorporating trophic complexity. Ecol Lett 10:522–538. Dunne, J. A., R. J. Williams, and N. D. Martinez, 2002. Network structure and biodiversity loss in food webs: robustness increases with connectance. Ecol Lett 5:558–567. Emmerson, M. and J. M. Yearsley, 2004. Weak interactions, omnivory and emergent food-web properties. Proc R Soc Lond B 271:397–405. Espinoza, R. E., J. J. Wiens, and C. R. Tracy, 2004. Recurrent evolution of herbivory in small, cold-climate lizards: breaking the ecophysiological rules of reptilian herbivory. Proc Natl Acad Sci USA 101:16819–16824. Essington, T. E., A. H. Beaudreau, and J. Wiedenmann, 2006. Fishing through marine food webs. Proc Natl Acad Sci USA 103:3171–3175. Estes, S. and S. J. Arnold, 2007. Resolving the paradox of stasis: models with stabilizing selection explain evolutionary divergence on all timescales. Am Nat 169:227–244. Fagan, W. F., 1997. Omnivory as a stabilizing feature of natural communities. Am Nat 150:554–567. Fagan, W. F., E. Siemann, C. Mitter, R. F. Denno, A. F. Huberty, H. A. Woods, and J. J. Elser, 2002. Nitrogen in insects: implications for trophic complexity and species diversification. Am Nat 160:784–802. Felsenstein, J., 1985. Phylogenies and the comparative method. Am Nat 125:1–15.  81  ———, 2008. Comparative methods with sampling error and within-species variation: contrasts revisited and revised. Am Nat 171:713–725. Finke, D. L. and R. F. Denno, 2006. Spatial refuge from intraguild predation: implications for prey suppression and trophic cascades. Oecologia 149:265–275. FitzJohn, R. G., 2010. Quantitative traits and diversification. Syst Biol 59:619–633. Frank, K. T., B. Petrie, J. S. Choi, and W. C. Leggett, 2005. Trophic cascades in a formerly cod-dominated ecosystem. Science 308:1621–1623. Freckleton, R. P., P. H. Harvey, and M. Pagel, 2002. Phylogenetic analysis and comparative data: a test and review of evidence. Am Nat 160:712–726. Froese, R. and D. Pauly, 2011. FishBase. World Wide Web electronic publication. www.fishbase.org, version (02/2011). Garland Jr, T. and A. R. Ives, 2000. Using the past to predict the present: confidence intervals for regression equations in phylogenetic comparative methods. Am Nat 155:346–364. Gavrilets, S. and A. Vose, 2005. Dynamic patterns of adaptive radiation. Proc Natl Acad Sci USA 102:18040–18045. Glor, R. E., J. J. Kolbe, R. Powell, A. Larson, and J. B. Losos, 2003. Phylogenetic analysis of ecological and morphological diversification in Hispaniolan trunkground anoles (Anolis cybotes group). Evolution 57:2383–2397. Goldwasser, L. and J. Roughgarden, 1993. Construction and analysis of a large Caribbean food web. Ecology 74:1216–1233. Grant, P. R. and B. R. Grant, 2006. Evolution of character displacement in Darwin’s finches. Science 313:224–226. Guill, C. and B. Drossel, 2008. Emergence of complexity in evolving niche-model food webs. J Theor Biol 251:108–120. Hairston Jr, N. G., S. P. Ellner, M. A. Geber, T. Yoshida, and J. A. Fox, 2005. Rapid evolution and the convergence of ecological and evolutionary time. Ecol Lett 8:1114–1127. Hallacher, L. E. and D. A. Roberts, 1985. Differential utilization of space and food by the inshore rockfishes (Scorpaenidae: Sebastes) of Carmel Bay, California. Environ Biol Fish 12:91–110. Hansen, T. F., 1997. Stabilizing selection and the comparative analysis of adaptation. Evolution 51:1341–1351. Hansen, T. F. and E. P. Martins, 1996. Translating between microevolutionary process and macroevolutionary patterns: the correlation structure of interspecific data. Evolution 50:1404–1417. Harmon, L., J. Weir, C. D. Brock, R. E. Glor, and W. Challenger, 2008. GEIGER: investigating evolutionary radiations. Bioinformatics 24:129–131. 82  Harmon, L. J. and J. B. Losos, 2005. The effect of intraspecific sample size on type I and type II error rates in comparative studies. Evolution 59:2705–2710. Harmon, L. J., J. B. Losos, T. J. Davies, R. G. Gillespie, J. L. Gittleman, W. B. Jennings, K. H. Kozak, M. A. McPeek, F. Moreno-Roark, T. J. Near, A. Purvis, R. E. Ricklefs, D. Schluter, J. A. Schulte II, O. Seehausen, B. L. Sidlauskas, O. Torres-Carvajal, J. T. Weir, and A. Ø. Mooers, 2010. Early bursts of body size and shape evolution are rare in comparative data. Evolution 64:2385–2396. Harmon, L. J., B. Matthews, S. Des Roches, J. M. Chase, J. B. Shurin, and D. Schluter, 2009. Evolutionary diversification in stickleback affects ecosystem functioning. Nature 458:1167–1170. Harvey, C. J., N. Tolimieri, and P. S. Levin, 2006. Changes in body size, abundance, and energy allocation in rockfish assemblages of the northeast Pacific. Ecol Appl 16:1502–1515. Heithaus, M. R., A. Frid, A. J. Wirsing, and B. Worm, 2008. Predicting ecological consequences of marine top predator declines. Trends Ecol Evol 23:202–210. Helvey, M., 1982. First observations of courtship behavior in rockfish, genus Sebastes. Copeia 1982:763–770. Hendrixson, H. A., R. W. Sterner, and A. D. Kay, 2007. Elemental stoichiometry of freshwater fishes in relation to phylogeny, allometry and ecology. J Fish Biol 70:121–140. Holt, R. D. and G. R. Huxel, 2007. Alternative prey and the dynamics of intraguild predation: theoretical perspectives. Ecology 88:2706–2712. Holt, R. D. and G. A. Polis, 1997. A theoretical framework for intraguild predation. Am Nat 149:745–764. Holyoak, M. and S. Sachdev, 1998. Omnivory and the stability of simple food webs. Oecologia 117:413–419. Hulsey, C. D., 2006. Function of a key morphological innovation: fusion of the cichlid pharyngeal jaw. Proc R Soc Lond B 273:669–675. Hyde, J. R., C. A. Kimbrell, J. E. Budrick, E. A. Lynn, and R. D. Vetter, 2008. Cryptic speciation in the vermilion rockfish (Sebastes miniatus) and the role of bathymetry in the speciation process. Mol Ecol 17:1122–1136. Hyde, J. R. and R. D. Vetter, 2007. The origin, evolution, and diversification of rockfishes of the genus Sebastes (Cuvier). Mol Phylogenet Evol 44:790–811. Ingram, T., 2011. Speciation along a depth gradient in a marine adaptive radiation. Proc R Soc Lond B 278:613–618. Ingram, T., L. J. Harmon, and J. B. Shurin, 2009. Niche evolution, trophic structure, and species turnover in model food webs. Am Nat 174:56–67.  83  Ingram, T., B. Matthews, C. Harrod, T. Stephens, J. Grey, R. Markel, and A. Mazumder, 2007. Lipid extraction has little effect on the δ 15 N of aquatic consumers. Limnol Oceanogr: Methods 5:338–343. Ingram, T. and J. B. Shurin, 2009. Trait-based assembly and phylogenetic structure in northeast Pacific rockfish assemblages. Ecology 90:2444–2453. Ingram, T. and M. Steel, 2010. Modelling the unpredictability of future biodiversity in ecological networks. J Theor Biol 264:1047–1056. Jennings, S., J. K. Pinnegar, N. V. C. Polunin, and T. W. Boon, 2001. Weak cross-species relationships between body size and trophic level belie powerful size-based trophic structuring in fish communities. J Anim Ecol 70:934–944. Johansson, M. L., M. A. Banks, K. D. Glunt, H. M. Hassel-Finnegan, and V. P. Buonaccorsi, 2008. Influence of habitat discontinuity, geographical distance, and oceanography on finescale population genetic structure of copper rockfish (Sebastes caurinus). Mol Ecol 17:3051–3061. Johns, G. C. and J. C. Avise, 1998. Tests for ancient species flocks based on molecular phylogenetic appraisals of Sebastes rockfishes and other marine fishes. Evolution 52:1135–1146. Johnson, M. and J. Stinchcombe, 2007. An emerging synthesis between community ecology and evolutionary biology. Trends Ecol Evol 22:250–257. Jonsson, B., 2001. Polymorphism and speciation in Arctic charr. J Fish Biol 58:605–638. Kassen, R., 2002. The experimental evolution of specialists, generalists, and the maintenance of diversity. J Evol Biol 15:173–190. Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli, 2001. The strength of phenotypic selection in natural populations. Am Nat 157:245–261. Kleiber, M., 1947. Body size and metabolic rate. Physiol Rev 27:511–541. Knowlton, N., L. A. Weigt, L. A. Sol´orzano, D. K. Mills, and E. Bermingham, 1993. Divergence in proteins, mitochondrial DNA, and reproductive compatibility across the isthmus of Panama. Science 260:1629–1632. Kocher, T. D., 2004. Adaptive evolution and explosive speciation: the cichlid fish model. Nature Rev Genet 5:288–298. Kondoh, M., 2003. Foraging adaptation and the relationship between food-web complexity and stability. Science 299:1388–1391. ———, 2008. Building trophic modules into a persistent food web. Proc Natl Acad Sci USA 105:16631–16635. Kratina, P., E. Hammill, and B. R. Anholt, 2010. Stronger inducible defences enhance persistence of intraguild prey. J Anim Ecol 79:993–999. 84  Langerhans, R. B., C. A. Layman, A. M. Shokrollahi, and T. J. DeWitt, 2004. Predator-driven phenotypic diversification in Gambusia affinis. Evolution 58:2305–2318. Lavin, P. A. and J. D. McPhail, 1986. Adaptive divergence of trophic phenotype among freshwater populations of the threespine stickleback (Gasterosteus aculeatus). Can J Fish Aquat Sci 43:2455–2463. Layman, C. A., K. O. Winemiller, D. A. Arrington, and D. B. Jepsen, 2005. Body size and trophic position in a diverse tropical food web. Ecology 86:2530–2535. Lea, R. N. and J. N. Fitch, 1972. Sebastes rufinanus, a New Scorpaenid Fish from Californian Waters. Copeia 1972:423–427. Legendre, P. and L. Legendre, 1998. Numerical ecology, 2nd edn. 2nd ed. Elsevier, Amsterdam. Levine, S., 1980. Several measures of trophic structure applicable to complex food webs. J Theor Biol 83:195–207. Lindeman, R. L., 1942. The trophic-dynamic aspect of ecology. Ecology 23:399–417. Loeuille, N. and M. Loreau, 2005. Evolutionary emergence of size-structured food webs. Proc Natl Acad Sci USA 102:5761–5766. Losos, J. B., T. W. Schoener, and D. A. Spiller, 2004. Predator-induced behaviour shifts and natural selection in field-experimental lizard populations. Nature 432:505–508. Love, M. S., M. Yoklavich, and L. K. Thorsteinson, 2002. The rockfishes of the northeast Pacific. University of California Press, Berkeley, CA. Mangel, M., H. K. Kindsvater, and M. B. Bonsall, 2007. Evolutionary analysis of life span, competition, and adaptive radiation, motivated by the pacific rockfishes (Sebastes). Evolution 61:1208–1224. Matthews, B., K. B. Marchinko, D. I. Bolnick, and A. Mazumder, 2010. Specialization of trophic position and habitat use by sticklebacks in an adaptive radiation. Ecology 91:1025–1034. May, R. M., 1973. Stability and complexity in model ecosystems. Princeton University Press, Princeton, NJ. McCann, K. S. and A. Hastings, 1997. Re-evaluating the omnivory-stability relationship in food webs. Proc R Soc Lond B 264:1249–1254. McCann, K. S., A. Hastings, and G. R. Huxel, 1998. Weak trophic interactions and the balance of nature. Nature 395:794–798. McCauley, E., 1984. The estimation of the abundance and biomass of zooplankton in samples. . Pp. 228–265, in J. A. Downing and F. H. Rigler, eds. A manual on methods for the assessment of secondary productivity in fresh waters. Blackwell Scientific, Oxford, UK. McKane, A. J., 2004. Evolving complex food webs. Eur Phys J B 38:287–295.  85  McPeek, M. A., 2008. The ecological dynamics of clade diversification and community assembly. Am Nat 172:E270–E284. McPhail, J. D., 1994. Speciation and the evolution of reproductive isolation in the sticklebacks (Gasterosteus) of south-western British Columbia. Pp. 399–437, in M. A. Bell and S. A. Foster, eds. The evolutionary biology of the threespine stickleback. Oxford University Press, Oxford, UK. Milinski, M. and R. Heller, 1978. Influence of a predator on the optimal foraging behaviour of sticklebacks (Gasterosteus aculeatus L.). Nature 275:642–644. Mintenbeck, K., U. Jacob, R. Knust, W. E. Arntz, and T. Brey, 2007. Depth-dependence in stable isotope ratio δ 15 N of benthic POM consumers: The role of particle dynamics and organism trophic guild. Deep Sea Res I 54:1015–1023. Moodie, G. E. E., 1972. Predation, natural selection and adaptation in an unusual threespine stickleback. Heredity 28:155–167. Morin, P., 1999. Productivity, intraguild predation, and population dynamics in experimental food webs. Ecology 80:752–760. Myers, R. A., J. K. Baum, T. D. Shepherd, S. P. Powers, and C. H. Peterson, 2007. Cascading effects of the loss of apex predatory sharks from a coastal ocean. Science 315:1846. Nee, S., 2006. Birth-death models in macroevolution. Annu Rev Ecol Evol Syst 37:1–17. Nee, S., E. C. Holmes, R. M. May, and P. H. Harvey, 1994. Extinction rates can be estimated from molecular phylogenies. Phil Trans R Soc Lond B 344:77–82. Nosil, P. and B. J. Crespi, 2006. Experimental evidence that predation promotes divergence in adaptive radiation. Proc Natl Acad Sci USA 103:9090–9095. Oksanen, L., S. D. Fretwell, and O. Jarvinen, 1979. Interspecific aggression and the limiting similarity of close competitors: the problem of size gaps in some community arrays. Am Nat 114:117–129. O’Meara, B. C., C. An´e, M. J. Sanderson, and P. C. Wainwright, 2006. Testing for different rates of continuous trait evolution using likelihood. Evolution 60:922–933. Ostrom, N. E., S. A. Macko, D. Deibel, and R. J. Thompson, 1997. Seasonal variation in the stable carbon and nitrogen isotope biogeochemistry of a coastal cold ocean environment. Geochim Cosmochim Acta 61:2929–2942. Pace, M. L., J. J. Cole, S. R. Carpenter, and J. F. Kitchell, 1999. Trophic cascades revealed in diverse ecosystems. Trends Ecol Evol 14:483–488. Pagel, M., 1997. Inferring evolutionary processes from phylogenies. Zool Scripta 26:331–348. Palkovacs, E. P., M. C. Marshall, B. A. Lamphere, B. R. Lynch, D. J. Weese, D. F. Fraser, D. N. Reznick, C. M. Pringle, and M. T. Kinnison, 2009. Experimental evaluation of evolution and coevolution as agents of ecosystem change in Trinidadian streams. Phil Trans R Soc Lond B 364:1617–1628. 86  Palkovacs, E. P. and D. M. Post, 2009. Experimental evidence that phenotypic divergence in predators drives community divergence in prey. Ecology 90:300–305. Palumbi, S. R., 1994. Genetic divergence, reproductive isolation, and marine speciation. Annu Rev Ecol Syst 25:547–572. Paradis, E., J. Claude, and K. Strimmer, 2004. APE: analyses of phylogenetics and evolution in R language. Bioinformatics 20:289–290. Parker, S. J., S. A. Berkeley, J. T. Golden, D. R. Gunderson, J. Heiretz, M. A. Hixon, R. Larson, B. M. Leaman, M. S. Love, J. A. Musick, V. M. O’Connell, S. Ralston, H. J. Weeks, and M. M. Yoklavich, 2000. Management of Pacific rockfish. Fisheries 25:22–29. Pauly, D. and V. Christensen, 1995. Primary production required to sustain global fisheries. Nature 374:255–257. Pauly, D., V. Christensen, J. Dalsgaard, R. Froese, and F. Torres, 1998. Fishing down marine food webs. Science 279:860–863. Persson, L., A. M. De Roos, D. Claessen, P. Bystrom, J. Lovgren, S. Sjogren, R. Svanb¨ack, E. Wahlstrom, and E. Westman, 2003. Gigantic cannibals driving a whole-lake trophic cascade. Proc Natl Acad Sci USA 100:4035–4039. Peters, R. H., 1983. The ecological implications of body size. Cambridge University Press, Cambridge, UK. Phillimore, A. B. and T. D. Price, 2008. Density-dependent cladogenesis in birds. PLoS Biol 6:e71. Pickett, S. T. A. and F. A. Bazzaz, 1978. Organization of an assemblage of early successional species on a soil-moisture gradient. Ecology 59:1248–1255. Pimm, S. L., 1980. Properties of food webs. Ecology 61:219–225. ———, 1982. Food webs. Chapman and Hall, London, UK. Pimm, S. L. and J. H. Lawton, 1978. On feeding on more than one trophic level. Nature 275:542–544. Polis, G. A., 1991. Complex trophic interactions in deserts: an empirical critique of food-web theory. Am Nat 138:123–155. Polis, G. A. and R. D. Holt, 1992. Intraguild predation: The dynamics of complex trophic interactions. Trends Ecol Evol 7:151–154. Polis, G. A., C. A. Myers, and R. D. Holt, 1989. The ecology and evolution of intraguild predation: potential competitors that eat each other. Annu Rev Ecol Syst 20:297–330. Post, D. M., 2002a. The long and short of food-chain length. Trends Ecol Evol 17:269–277. ———, 2002b. Using stable isotopes to estimate trophic position: models, methods, and assumptions. Ecology 83:703–718. 87  Post, D. M., C. A. Layman, D. A. Arrington, G. Takimoto, J. Quattrochi, and C. G. Monta˜na, 2007. Getting to the fat of the matter: models, methods and assumptions for dealing with lipids in stable isotope analyses. Oecologia 152:179–189. Post, D. M. and E. P. Palkovacs, 2009. Eco-evolutionary feedbacks in community and ecosystem ecology: interactions between the ecological theatre and the evolutionary play. Phil Trans R Soc Lond B 364:1629–1640. Post, D. M. and G. Takimoto, 2007. Proximate structural mechanisms for variation in foodchain length. Oikos 116:775–782. Pressley, P. H., 1981. Parental effort and the evolution of nest-guarding tactics in the threespine stickleback, Gasterosteus aculeatus L. Evolution 35:282–295. Pritchard, J. R. and D. Schluter, 2001. Declining interspecific competition during character displacement: summoning the ghost of competition past. Evol Ecol Res 3:209–220. Puebla, O., 2009. Ecological speciation in marine v. freshwater fishes. J Fish Biol 75:960–996. Pybus, O. G. and P. H. Harvey, 2000. Testing macro-evolutionary models using incomplete molecular phylogenies. Proc R Soc Lond B 267:2267–2272. R Development Core Team, 2009. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rabosky, D. L., 2006. LASER: a maximum likelihood toolkit for detecting temporal shifts in diversification rates from molecular phylogenies. Evol Bioinform Online 2:247. ———, 2010. Extinction rates should not be estimated from molecular phylogenies. Evolution 64:1816–1824. Rabosky, D. L. and I. J. Lovette, 2008. Density-dependent diversification in North American wood warblers. Proc R Soc Lond B 275:2363–2371. Reimchen, T. E., 1991. Trout foraging failures and the evolution of body size in stickleback. Copeia 1991:1098–1104. ———, 1994. Predators and morphological evolution in threespine stickleback. Pp. 240–273, in M. A. Bell and S. A. Foster, eds. The evolutionary biology of the threespine stickleback. Oxford University Press, Oxford, UK. Revell, L. J., L. J. Harmon, and D. C. Collar, 2008. Phylogenetic signal, evolutionary process, and rate. Syst Biol 57:591–601. Rezende, E. L., E. M. Albert, M. A. Fortuna, and J. Bascompte, 2009. Compartments in a marine food web associated with phylogeny, body mass, and habitat structure. Ecol Letters 12:779–788. Reznick, D. and J. A. Endler, 1982. The impact of predation on life history evolution in Trinidadian guppies (Poecilia reticulata). Evolution 36:160–177.  88  Reznick, D. A., H. Bryga, and J. A. Endler, 1990. Experimentally induced life-history evolution in a natural population. Nature 346:357–359. Richman, A. D. and T. Price, 1992. Evolution of ecological differences in the Old World leaf warblers. Nature 355:817–821. Ritchie, M. E. and H. Olff, 1999. Spatial scaling laws yield a synthetic theory of biodiversity. Nature 400:557–560. Rocha, L. A., D. R. Robertson, J. Roman, and B. W. Bowen, 2005. Ecological speciation in tropical reef fishes. Proc R Soc Lond B Biol Sci 272:573–579. Rooney, N., K. S. McCann, and J. C. Moore, 2008. A landscape theory for food web architecture. Ecol Lett 11:867–881. Rossberg, A. G., H. Matsuda, T. Amemiya, and K. Itoh, 2005. An explanatory model for food-web structure and evolution. Ecol Complex 2:312–321. Roughgarden, J., 1972. Evolution of niche width. Am Nat 106:683–718. R¨uber, L., E. Verheyen, and A. Meyer, 1999. Replicated evolution of trophic specializations in an endemic cichlid fish lineage from Lake Tanganyika. Proc Natl Acad Sci USA 96:10230–10235. Rundle, H. D., S. M. Vamosi, and D. Schluter, 2003. Experimental test of predation’s effect on divergent selection during character displacement in sticklebacks. Proc Natl Acad Sci USA 100:14943–14948. Sample, B. E., R. J. Cooper, R. D. Greer, and R. C. Whitmore, 1993. Estimation of insect biomass by length and width. Am Midl Nat 129:234–240. Sato, A., C. O’hUigin, F. Figueroa, P. R. Grant, B. R. Grant, H. Tichy, and J. Klein, 1999. Phylogeny of Darwin’s finches as revealed by mtDNA sequences. Proc Natl Acad Sci USA 96:5101–5106. Schluter, D., 1994. Experimental evidence that competition promotes divergence in adaptive radiation. Science 266:798–801. ———, 2000. The ecology of adaptive radiation. Oxford University Press, Oxford, UK. ———, 2001. Ecology and the origin of species. Trends Ecol Evol 16:372–380. Schluter, D., E. A. Clifford, M. Nemethy, and J. S. McKinnon, 2004. Parallel evolution and inheritance of quantitative traits. Am Nat 163:809–822. Schluter, D. and J. D. McPhail, 1992. Ecological character displacement and speciation in sticklebacks. Am Nat 140:85–108. Schoener, T. W., 1989. Food webs from the small to the large: the Robert H. MacArthur award lecture. Ecology 70:1559–1589. ———, 2011. The newest synthesis: understanding the interplay of evolutionary and ecological dynamics. Science 331:426–429. 89  Seehausen, O., Y. Terai, I. S. Magalhaes, K. L. Carleton, H. D. J. Mrosso, R. Miyagi, I. van der Sluijs, M. V. Schneider, M. E. Maan, H. Tachida, H. Imai, and N. Okada, 2008. Speciation through sensory drive in cichlid fish. Nature 455:620–626. Shurin, J. B., D. S. Gruner, and H. Hillebrand, 2006. All wet or dried up? Real differences between aquatic and terrestrial food webs. Proc R Soc Lond B 273:1–9. Siemers, B. M., S. Greif, I. Borissov, S. L. Voigt-Heucke, and C. C. Voigt, 2011. Divergent trophic levels in two cryptic sibling bat species. Oecologia 166:69–78. Simpson, G. G., 1953. The major features of evolution. Columbia University Press, New York, NY. Sivasundar, A. and S. R. Palumbi, 2010. Parallel amino acid replacements in the rhodopsins of the rockfishes (Sebastes spp.) associated with shifts in habitat depth. J Evol Biol 23:1159–1169. Sk´ulason, S. and T. B. Smith, 1995. Resource polymorphisms in vertebrates. Trends Ecol Evol 10:366–370. Slatkin, M., 1980. Ecological character displacement. Ecology 61:163–177. Spoljaric, M. A. and T. E. Reimchen, 2007. 10 000 years later: evolution of body shape in Haida Gwaii threespined stickleback. J Fish Biol 70:1484–1503. Srivastava, D. S. and T. Bell, 2009. Reducing horizontal and vertical diversity in a foodweb triggers extinctions and impacts functions. Ecol Lett 12:1016–1028. Tabachnick, B. G. and L. S. Fidell, 2001. Using multivariate statistics, 3rd edn. Harper and Row, New York, NY. Taper, M. L. and T. J. Case, 1985. Quantitative genetic models for the coevolution of character displacement. Ecology 66:355–371. Thompson, R. M., M. Hemberg, B. M. Starzomski, and J. B. Shurin, 2007. Trophic levels and trophic tangles: the prevalence of omnivory in real food webs. Ecology 88:612–617. Vamosi, S. M., 2003. The presence of other fish species affects speciation in threespine sticklebacks. Evol Ecol Res 5:717–730. Vamosi, S. M. and D. Schluter, 2002. Impacts of trout predation on fitness of sympatric sticklebacks and their hybrids. Proc R Soc Lond B 269:923–930. Van Doorn, G. S., P. Edelaar, and F. J. Weissing, 2009. On the origin of species by natural and sexual selection. Science 326:1704–1707. Vance-Chalcraft, H. D., J. A. Rosenheim, J. R. Vonesh, C. W. Osenberg, and A. Sih, 2007. The influence of intraguild predation on prey suppression and prey release: a meta-analysis. Ecology 88:2689–2696. Vander Zanden, M. J., G. Cabana, and J. B. Rasmussen, 1997. Comparing trophic position of freshwater fish calculated using stable nitrogen isotope ratios (δ 15 N) and literature dietary data. Can J Fish Aquat Sci 54:1142–1158. 90  Vander Zanden, M. J. and W. W. Fetzer, 2007. Global patterns of aquatic food chain length. Oikos 116:1378–1388. Vander Zanden, M. J. and J. B. Rasmussen, 1996. A trophic position model of pelagic food webs: impact on contaminant bioaccumulation in lake trout. Ecol Monogr 66:451–477. Vanderklift, M. A. and S. Ponsard, 2003. Sources of variation in consumer-diet δ 15 N enrichment: a meta-analysis. Oecologia 136:169–182. Vandermeer, J., 2006. Omnivory and the stability of food webs. J Theor Biol 238:497–504. Venables, W. N. and B. D. Ripley, 2002. Modern applied statistics with S, 4th edn. Springer, New York, NY. Wagenmakers, E.-J. and S. Farrell, 2004. AIC model selection using Akaike weights. Psychon Bull Rev 11:192–196. Wagner, C. E., P. B. McIntyre, K. S. Buels, D. M. Gilbert, and E. Michel, 2009. Diet predicts intestine length in Lake Tanganyika’s cichlid fishes. Funct Ecol 23:1122–1131. Wainwright, P. C., D. R. Bellwood, M. W. Westneat, J. R. Grubich, and A. S. Hoey, 2004. A functional morphospace for the skull of labrid fishes: patterns of diversity in a complex biomechanical system. Biol J Linn Soc 82:1–25. Walker, J. A., 1997. Ecological morphology of lacustrine threespine stickleback Gasterosteus aculeatus L. (Gasterosteidae) body shape. Biol J Linn Soc 61:3–50. ———, 2000. Net evolutionary trajectories of body shape evolution within a microgeographic radiation of threespine sticklebacks (Gasterosteus aculeatus). J Zool 252:293–302. Warrant, E. J. and N. A. Locket, 2004. Vision in the deep sea. Biol Rev 79:671–712. Werner, E. E. and D. J. Hall, 1988. Ontogenetic habitat shifts in bluegill: the foraging rate-predation risk trade-off. Ecology 69:1352–1366. Williams, R. J. and N. D. Martinez, 2000. Simple rules yield complex food webs. Nature 404:180–183. ———, 2004. Limits to trophic levels and omnivory in complex food webs: theory and data. Am Nat 163:458–468. Yoshida, K., 2003. Dynamics of evolutionary patterns of clades in a food web system model. Ecol Res 18:625–637. Yoshida, T., L. E. Jones, S. P. Ellner, G. F. Fussmann, and N. G. Hairston, 2003. Rapid evolution drives ecological dynamics in a predator-prey system. Nature 424:303–306. Zelditch, M. L., D. L. Swiderski, H. D. Sheets, and W. L. Fink, 2004. Geometric morphometrics for biologists. Elsevier Academic Press, Amsterdam, Netherlands.  91  Appendix A  Sources of rockfish specimens Species  Common Name  n  Sample sources  S. aleutianus  Rougheye Rockfish  5  BC-62-424, UW-113740, HG  S. alutus  Pacific Ocean Perch  10  HG  Kelp Rockfish  7 (2)  UW-114048, UW-047421,  S. atrovirens  SIO-63-1024-53D S. auriculatus  Brown Rockfish  12 (5)  SB  S. aurora  Aurora Rockfish  7  BC-67-12, BC-70-36, HG  S. babcocki  Redbanded Rockfish  7  HG  S. borealis  Shortraker Rockfish  5  UW-111790, UW-114735, UW-113747, HG  S. brevispinis  Silvergray Rockfish  5  HG  S. carnatus  Gopher Rockfish  6 (5)  SB  S. caurinus  Copper Rockfish  45 (14)  BC-18, BC-19, BS, SB  S. chlorostictus  Greenspotted Rockfish  7 (2)  UW-043412, UW-043407, SB  S. chrysomelas  Black and Yellow Rockfish  5 (1)  UW-002755, UW-114053, SIO-H47-201, SIO-H48-315, SB  S. ciliatus  Dark Rockfish  5  BC-65-791, UW-046483, UW-014416, UW-004776, SIO-94-179  Starry Rockfish  7 (2)  UW-114066, UW-113199, UW-022422, SB  Darkblotched Rockfish  7  BC-63-924, BC-65-789  Calico Rockfish  7 (5)  SB  S. diploproa  Splitnose Rockfish  7  BC-70-33, BC-63-924, BC-70-37, HG  S. elongatus  Greenstriped Rockfish  11 (4)  HG, SB  S. emphaeus  Puget Sound Rockfish  7  BC-66-135  S. ensifer  Swordspine Rockfish  7 (5)  SB  S. constellatus S. crameri S. dallii  92  Species  Common Name  n  Sample sources  S. entomelas  Widow Rockfish  9 (5)  HG, SB  Pink Rockfish  4 (2)  SIO-68-142, SB  Yellowtail Rockfish  11 (2)  BS, HG, SB  Bronzespotted Rockfish  3  SIO-322-488, SIO-68-259-53, SIO-97-58  Chilipepper Rockfish  9 (7)  SB  S. helvomaculatus  Rosethorn Rockfish  7  HG  S. hopkinsi  Squarespot Rockfish  7  UW-040747, UW-113587, UW-113588,  S. eos S. flavidus S. gilli S. goodei  UW-114026, UW-21426 Shortbelly Rockfish  7  BC-06-0171  S. lentiginosus  Freckled Rockfish  3  SIO-65-6-53  S. levis  Cowcod Rockfish  8  UW-40093, UW-114067, UW-040227,  S. jordani  UW-043414, UW-047548 S. macdonaldi  Mexican Rockfish  4 (1)  UW-114065, SIO-65-64, SB  S. maliger  Quillback Rockfish  6 (6)  BS  Black Rockfish  32 (17)  BS  S. melanosema  Semaphore Rockfish  2  UW-112698, SIO-06-85  S. melanostictus  Blackspotted Rockfish  5  BC-62-720, UW-114746,  S. melanops  UW-049336, SIO-88-187, HG S. melanostomus  Blackgill Rockfish  7  UW-113590, UW-113458, UW-043501, UW-043479, UW-043501  Vermillion Rockfish  14 (10)  SB  Whitespeckled Rockfish  1  SIO-95-33  S. mystinus  Blue Rockfish  19 (7)  SB  S. nebulosus  China Rockfish  7 (1)  BC-H11, BC-55-309, BC-65-373  S. nigrocinctus  Tiger Rockfish  7  BC-59-168, BC-63-938, BC-63-934  Speckled Rockfish  4  UW-49713, UW-114056,  S. miniatus S. moseri  S. ovalis  UW-113586, SIO-67-9-53 Boccaccio  7 (6)  HG, SB  S. phillipsi  Chameleon Rockfish  4 (1)  BC-64-476, SIO-02-14SB  S. pinniger  Canary Rockfish  15 (10)  BS, HG  S. polyspinis  Northern Rockfish  7  BC-65-148, BC-62-508, BC-65-78  S. proriger  Redstripe Rockfish  8  HG  Grass Rockfish  5  UW-114047, UW-018991,  S. paucispinis  S. rastrelliger  UW-047283, SIO-H49-68, SIO-63-1056 S. reedi  Yellowmouth Rockfish  6  93  HG  Species  Common Name  n  Sample sources  S. rosaceus  Rosy Rockfish  4  UW-114046, UW-002865, SIO-68-182-53  S. rosenblatti  Greenblotched Rockfish  8 (5)  SB  S. ruberrimus  Yelloweye Rockfish  9 (6)  BC-74-25, BS  S. rubrivinctus  Flag Rockfish  6 (3)  SB  Dwarf-red Rockfish  2  see below  Bank Rockfish  12 (10)  SB  Stripetail Rockfish  7 (2)  BC-70-32, UW-047442, SB  S. semicinctus  Halfbanded Rockfish  7  SB  S. serranoides  Olive Rockfish  8 (5)  SB  S. serriceps  Treefish  10 (8)  SB  S. simulator  Pinkrose Rockfish  11 (5)  SB  S. umbrosus  Honeycomb Rockfish  7  UW-29694, UW-022425,  S. rufinanus S. rufus S. saxicola  UW-022424, UW-114032 S. variabilis  Dusky Rockfish  6  BC-65-148, BC-62-676  S. variegatus  Harlequin Rockfish  7  BC-62-771, BC-64-292  Pygmy Rockfish  7  BC-64-275, BC-66-137, BC-66-131  Sharpchin Rockfish  8  HG  S. wilsoni S. zacentrus  Sample sizes (n) are shown for morphological data, with sample sizes for stable isotope analysis in parentheses where applicable. Accession numbers correspond to the following collections: UW = University of Washington Fish Collection; BC = University of British Columbia Fish Museum; SIO = Scripps Institute of Oceanography Marine Vertebrate Collection. Other labels correspond to fish obtained from Department of Fisheries and Oceans Canada trawl surveys off the west coast of Haida Gwaii, British Columbia (HG), and from recreational anglers in Barkley Sound, British Columbia (BS) and the Santa Barbara Channel, California (SB). Measurements for S. rufinanus were taken from the original species description (Lea and Fitch, 1972).  94  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items