Community ecology in a warmer world: direct and indirect effects of temperature on community dynamics

by

Rebecca Lee Kordas

B.A., The University of Chicago 2002 M.Sc., California State University Northridge 2006

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

in

The Faculty of Graduate and Postdoctoral Studies

(Zoology)

THE UNIVERSITY OF BRITISH COLUMBIA

(Vancouver)

September 2014

© Rebecca Lee Kordas 2014

Abstract

As the Earth’s climate changes so too do its ecosystems, due to shifts in abundance, biodiversity and interaction strengths among their constituent species. Although warming will simultaneously affect many aspects of ecological communities, disentangling the abiotic and biotic contributions will improve our understanding of how assemblages of interacting species will respond to climate change. My goal was to determine how warming affects community assemblages via direct (mediated by organismal physiology) vs. indirect effects (mediated by species interactions). I addressed this with 12-16 month-long manipulative experiments in the rocky intertidal zone of Salt Spring Island, Canada. I created a novel in situ method for increasing substratum temperature for settling benthic organisms, using black- and white-bordered settlement plates. In the first experiment (Chapter 3), I monitored the response of functional groups and diversity to warmed treatments. Results from this experiment suggest that communities in thermally stressful habitats respond to warming via the interplay between species-specific physiological responses and secondary adaptive strategies such as behavioral microhabitat selection. In Chapter 4, I concentrated on the direct effects of warming. As a case study, I monitored the direct effects of in situ warming on the vital rates of two competing barnacle species. Warming negatively affected both species of barnacles, however the population of the competitive dominant was more severely affected than the subordinate species, leading to a temperature-induced change in space occupancy. In Chapter 5, I focused on the indirect effects of warming on community dynamics by manipulating temperature and herbivore access to communities. Community structure and successional trajectory differed markedly between treatments, due to disturbance from herbivores and high species turnover due to warming. Despite the stochastic nature of development, warmed communities with herbivores ultimately lost the variability created by herbivore-associated disturbances, resulting in highly similar assemblages between warm and cool treatments. These results illustrate how environmental change can alter species-specific thermal responses, complex population dynamics, and interaction strengths, with cascading impacts on community dynamics. They further demonstrate how assemblages of multiple, interacting species will respond to climate change, which is imperative if we hope to effectively prepare for and adapt to its effects.

ii Preface

Some of the work included in this dissertation has been published or submitted for publication to peer-reviewed scientific journals: A version of Chapter 2 has been published and the appropriate license has been obtained from Elsevier. I was the lead investigator although all authors contributed text to the manuscript:

Kordas, RL, CDG Harley, MI O’Connor. 2011. Community ecology in a warming world: the influence of temperature on interspecific interactions in marine systems. Journal of Experimental Marine Biology and Ecology 400(1-2): 218-226.

A version of Chapter 3 has been submitted for publication. On March 12, the manuscript was provisionally accepted pending revisions, which were submitted on April 23. Another set of minor revisions were requested on June 7, which will be submitted in early September, 2014. I designed the experiment, conducted all field work, conducted initial data analyses, and wrote the manuscript. Stefan Storey assisted in development of the warming plate design and some field work. Steve Dudgeon analyzed the data using AIC methods. Chris Harley was the supervisory author on this project and was involved throughout the project in concept formation and manuscript edits.

Kordas, RL, S Dudgeon, S Storey, CDG Harley. Revision in review. Intertidal community responses to field-based experimental warming. Oikos.

Chapters 3 and 4 originated from the same experimental design, although different data and analyses were used for the manuscript. Accordingly, I designed the experiment, collected and analyzed all data, and wrote the manuscript for Chapter 4. Chris Harley was the supervisory author on this project and was involved throughout the project in concept formation and manuscript edits. Chapter 5 uses the same experimental design as Chapters 3-4, though it was conducted at a different time and location. I designed the experiment, conducted field work, collected and analyzed all data, and wrote the manuscript. Chris Harley was the supervisory author on this project and was involved throughout the project in concept formation and manuscript edits.

iii Table of Contents

Abstract ...... ii Preface ...... iii Table of Contents ...... iv List of Tables ...... vii List of Figures ...... ix Chapter 1 Introduction ...... 1 1.1 Community ecology: transitioning from a philosophical to an environmental agenda ...... 1 1.2 The rocky intertidal: a laboratory for global warming experiments ...... 2 1.3 Structure of this dissertation ...... 4 Chapter 2 The Influence of Temperature on Interspecific Interactions ...... 6 2.1 Synopsis ...... 6 2.2 Introduction ...... 6 2.3 The biological importance of temperature ...... 9 2.4 Interspecific variation in thermal sensitivity ...... 13 2.5 Incorporating time and space: phenology and biogeography ...... 17 2.6 The search for generality ...... 21 2.7 Future research priorities ...... 24 Chapter 3 Intertidal Community Responses to Field-Based Warming ...... 27 3.1 Synopsis ...... 27 3.2 Introduction ...... 27 3.3 Methods ...... 30 3.3.1 Study system ...... 30 3.3.2 Experimental warming treatment ...... 30 3.3.3 Statistical analyses ...... 33 3.4 Results ...... 35 3.4.1 Temperature ...... 35 3.4.2 Functional group responses ...... 38 3.4.3 Community response ...... 43 3.5 Discussion ...... 48 Chapter 4 Demographic Responses of Coexisting Species to In Situ Warming ...... 52 4.1 Synopsis ...... 52 4.2 Introduction ...... 53 4.3 Methods ...... 56 4.3.1 Study system ...... 56 4.3.2 Experimental warming treatment ...... 56 4.3.3 Statistical analyses ...... 58 4.4 Results ...... 60

iv 4.4.1 Temperature ...... 60 4.4.2 Barnacle abundance ...... 63 4.4.3 Barnacle survivorship ...... 73 4.4.4 Barnacle growth rate and size ...... 78 4.5 Discussion ...... 85 4.5.1 Efficacy of experimental temperature manipulation ...... 85 4.5.2 Lethal effects of warming ...... 86 4.5.3 Sublethal effects of warming ...... 88 4.5.4 Population responses and shifts in dominance ...... 89 4.5.5 Response to a changing climate ...... 90 4.5.6 Conclusions ...... 92 Chapter 5 Warming Modifies the Effect of Herbivory on Community Dynamics ...... 93 5.1 Synopsis ...... 93 5.2 Introduction ...... 94 5.3 Methods ...... 96 5.3.1 Study system ...... 96 5.3.2 Experimental warming treatment ...... 98 5.3.3 Limpet treatments ...... 98 5.3.4 Statistical analyses ...... 101 5.4 Results ...... 107 5.4.1 Treatment effectiveness ...... 107 5.4.2 Functional groups ...... 114 5.4.3 Species interactions ...... 116 5.4.4 Diversity ...... 118 5.4.5 Seasonal community structure ...... 120 5.4.6 Community trajectory ...... 125 5.5 Discussion ...... 129 5.5.1 Efficacy of warming and herbivore removals ...... 130 5.5.2 Independent effects of herbivores and warming on taxa ...... 131 5.5.3 Warming modifies species interactions ...... 133 5.5.4 Community structure ...... 135 5.5.5 Diversity ...... 138 5.5.6 Community trajectory ...... 138 5.5.7 Summary and implications ...... 140 Chapter 6 Concluding Remarks ...... 142 6.1 Synopsis ...... 142 6.2 Improving realism for global warming experiments ...... 144 6.2.1 Classic warming experiments ...... 144 6.2.2 Advantages and disadvantages of passive solar heated settlement plates ...... 145 6.2.3 Relevance of solar plates to climate warming ...... 147 6.3 Warming alters community dynamics via direct and indirect pathways ...... 149 6.4 Looking forward ...... 151 6.4.1 Global warming will be stressful ...... 151 6.4.2 Thermal dependency of species interactions ...... 152 6.4.3 Community and food web responses ...... 154

v Bibliography ...... 155 Appendix A Thermal Performance of Salt Spring Species ...... 180 Appendix B Linear Mixed Effect Model Results ...... 182 Appendix C Summary of Barnacle Thermal Responses ...... 185 Appendix D Taxonomic Responses to Warming and Limpets ...... 186 Appendix E Species Contributing to Differences Between Treatments ...... 188 Appendix F Species List for Salt Spring Island ...... 191

vi List of Tables

3.1 Summary statistics of plate temperatures from Salt Spring Island in 2009 ...... 38 3.2 Effects of temperature and zone on invertebrate and algae community structure ...... 45 3.3 Percentage contributions of invertebrate species to community differences ...... 46 3.4 Effects of temperature and zone on consumer community structure under plates ...... 47

4.1 Variance in temperature of plates and rock ...... 63 4.2 Response of barnacle density over one year ...... 67 4.3 Response of percent cover over one year ...... 68 4.4 Response of Balanus glandula density and percent cover ...... 71 4.5 Response of Chthamalus dalli density and percent cover ...... 72 4.6 Response of the proportion of dead barnacles ...... 74 4.7 Response of survivorship of young barnacle life stages ...... 75 4.8 Response of survivorship of newly metamorphosed barnacles only ...... 76 4.9 Response of survivorship of newly metamorphosed B. glandula and C. dalli ...... 78 4.10 Response of barnacle growth rates in summer and winter ...... 79 4.11 Response of B. glandula and C. dalli growth rates, separately ...... 80 4.12 Response of C. dalli growth rates ...... 83 4.13 Response of C. dalli basal area ...... 84

5.1 Angle and direction of plates installed in the rocky intertidal ...... 97 5.2 Analysis of mean temperature of plates ...... 108 5.3 Summary statistics of plate temperatures from Salt Spring Island in 2011-2012 ...... 109 5.4 Effect of temperature and limpet treatments on key taxa and diversity metrics ...... 112 5.5 Analysis of limpet abundance during low-tide and high-tide surveys ...... 113 5.6 Effect of treatments on community structure for summer 2011 ...... 122 5.7 Effect of treatments on community structure for winter 2012 ...... 122 5.8 Effect of treatments on community structure for summer 2012 ...... 123 5.9 Percentage contributions of species to community structure in summers and winter ....124 5.10 Effect of treatments on successional trajectories ...... 128

vii

A.1 Salt Spring algal species physiological thermal responses ...... 180 A.2 Salt Spring invertebrate species physiological thermal responses ...... 181

B.1 Linear mixed effect model analyses ...... 182

C.1 Mean effect of warming on B. glandula and C. dalli populations ...... 185

E.1 Species contributing to differences between treatments ...... 188

F.1 Species list for Salt Spring in 2011-2012 ...... 191

viii List of Figures

2.1 Projected surface temperature changes for the late 21st century...... 7 2.2 The pathway by which temperature influences individuals and populations...... 10 2.3 Relationship between temperature and various biological rates...... 11 2.4 Interspecific variation in the impact of rising temperatures...... 14 2.5 Direct and indirect effects of rising temperature on an interacting species pair...... 17 2.6 Hypothetical range shifts due to global warming...... 20

3.1 Passive solar heated plate design...... 32 3.2 Daily maximum temperature of plates and rock on Salt Spring Island in 2009 ...... 36 3.3 Plate, rock, and air temperature compared to tidal cycles over five days in August ...... 37 3.4 Mean residual daily maximum temperature of plates and rocky bench...... 37 3.5 Percent cover of all algae over one year...... 40 3.6 Percent cover of each algal functional group over one year...... 40 3.7 Density of all grazers over one year...... 41 3.8 Density of each grazer functional group over one year...... 41 3.9 Density of all barnacles over one year...... 42 3.10 Effect of the temperature treatment on species richness over one year...... 44 3.11 Community structure of invertebrates and algae on plate surfaces...... 45 3.12 Community structure of consumers under plates...... 47

4.1 Daily maximum temperature on plates and nearby rock on Salt Spring in 2009...... 62 4.2 Density of B. glandula and C. dalli over one year...... 65 4.3 Percent cover of B. glandula and C. dalli over one year...... 66 4.4 Representative photographs of black and white plates and the nearby rocky bench...... 69 4.5 Proportion of dead barnacles over one year...... 73 4.6 Mean survivorship of cyprids and newly metamorphosed barnacles...... 75 4.7 Survivorship of newly metamorphosed barnacles over two months...... 76 4.8 Mean daily growth rates of B. glandula and C. dalli...... 79 4.9 Chthamalus dalli mean daily growth rates and mean basal area over time...... 82

ix

5.1 Passive solar heated plate design with copper limpet-exclusions...... 99 5.2 Conceptual approach to interpreting the warming × limpet interaction...... 104 5.3 Residual daily maximum temperature of plates compared to nearby rock...... 108 5.4 Daily maximum temperature of plates and rock on Salt Spring Island in 2011-2012.....109 5.5 Density of limpets in treatments...... 111 5.6 Mean limpet density in each treatment during low and high-tide surveys...... 113 5.7 Mean abundance of common taxa in each treatment over 16 months...... 115 5.8 Mean abundance of B. glandula and diatoms in each treatment...... 117 5.9 Diversity in treatments over 16 months...... 119 5.10 Community structure for representative dates in summers and winter...... 121 5.11 Community trajectory in each treatment over 16 months...... 127 5.12 Variability in community trajectory in each treatment...... 129

6.1 The pathway by which temperature influences communities...... 143 6.2 Assumptions of using cooled treatments vs. warmed treatments...... 145 6.3 Interaction web for community in Salt Spring experiments...... 150 6.4 Thermal dependency of species interactions...... 153

D.1 Abundance of individual taxa in treatments over 16 months ...... 186

x Chapter 1 Introduction

1.1 Community ecology: transitioning from a philosophical to an environmental agenda

“One of the first things with which an ecologist has to deal is the fact that each different kind of habitat contains a characteristic set of animals. We call these animal associations, or better, animal communities, for we shall see later on that they are not mere assemblages of species living together, but form closely-knit communities or societies comparable to our own” (Elton 1927)

Ecological communities can be extraordinarily complex. For more than a century, ecologists have worked to understand this complexity by exploring the relationship between living organisms and their natural environment. These classic studies measured the abundance, distribution, and biomass of organisms as well as the nature of interactions between them (Wells 1928, Elton and Miller 1954). Modern community ecologists have progressed beyond the documentation of patterns, by using manipulative experiments to uncover the mechanisms underlying these patterns (Hairston 1989). However, some of the basic questions have yet to be fully answered, for example: Are the composition and structure of multispecies assemblages determined by interspecific interactions or adaptations of individual species? What is the relative importance of each, and how does the balance change over time, space, and environmental gradients? Is the answer context dependent or generalizable? Answers to these questions are no longer philosophical, but require real answers in the context of a changing climate. Anthropogenic carbon emission concentrations have reached record atmospheric levels and will continue to increase, affecting all components of the climate system (IPCC 2013). By 2100, global mean temperatures are expected to rise by 1.0-3.7°C over land and 0.6-2.0°C in surface oceans, causing changes to ocean circulation, an increase in the severity and frequency of extreme weather and thermal events, melting of glaciers, a 0.4-0.6 m rise in sea level, and changes in precipitation patterns (IPCC 2013). Climate change is undoubtedly one of the most serious threats facing earth’s biota (Thomas et al. 2004, Assessment 2005, Schroter 2005, Harley et al. 2006, Halpern et al. 2008, Pimm 2009, Hoegh-Guldberg and Bruno 2010), with extinction

1 estimates as high as 70% if global mean temperatures warm by 3.5°C (IPCC 2007). The vital next step is to understand the causes and consequences of climate change impacts in a general ecological framework so society can formulate predictions, reduce and quantify uncertainty, and respond to the effects of global environmental change.

1.2 The rocky intertidal: a laboratory for global warming experiments

Organisms in the intertidal zone have evolved to cope with extreme and fluctuating conditions. During low tide, organisms experience dryer and more variable terrestrial conditions, while at high tide, organisms experience more stable aqueous conditions. Tides also fluctuate over a bi- weekly cycle; extreme high and low (‘spring’) tides are associated with new and full moons, while the tidal range is at its minimum when the moon is at first or third quarter (termed ‘neap tides’). Over a year, environmental conditions can be more extreme when low tides coincide with extremely hot summer temperatures or extremely cold winter temperatures. Thus physical parameters such as temperature, humidity, and salinity can fluctuate drastically and quickly (Stickle and Denoux 1976, Denny and Harley 2006). Since all intertidal organisms are ectothermic species, their body temperature may vary by 20°C between high and low tides (Denny and Harley 2006). The precise location an organism inhabits in the intertidal is determined by a combination of abiotic and biotic conditions, such that species are often restricted to specific heights, or characteristic ‘zones’, in the intertidal. Nearly all producers and filter feeders are ‘sessile’ (stationary, like seaweeds and barnacles) while many ‘grazers’ (grazing herbivores) and carnivores are more mobile. However most of these consumers are slow moving (e.g. snails) and are unlikely to travel more than a couple meters in a day. Since marine conditions are thought to be more benign and space is a limited resource, biotic interactions (e.g. competition, predation) are often more intense in lower zones (Harley 2011), although exceptions to this general rule have been noted (Davison and Pearson 2008). Intertidal height is positively correlated with time exposed to terrestrial conditions, thus organisms higher up must be able to tolerate longer periods of ‘emersion’ (out of water). Many organisms in the intertidal live at or near their physiological limits for temperature and associated ‘desiccation’ (drying out) stress (Hofmann and Somero 1995, Somero 2002,

2 Davenport and Davenport 2005). Temperature affects nearly all biological processes and plays a pivotal role in governing the physiology and behavior of intertidal ectotherms. The relationship between temperature and physiological performance is unimodal; performance increases gradually with temperature up to some optimum, after which performance declines sharply (described in more detail in Chapter 2). Over the declining portion of the curve, organisms experience thermal stress which can induce the expression of heat shock proteins, cause oxidative damage, protein denaturation, and finally death (Pörtner 2010). The longer that intertidal organisms are exposed to sublethal stressful temperatures at low tide, the more metabolic resources are devoted to repairing and replacing heat damaged proteins (Somero 2002). Furthermore, species that live in the hottest intertidal habitats (i.e., in higher zones) have lethal thermal limits that closely match maximum environmental temperatures (Stillman and Somero 2000). To exacerbate the issue, some higher intertidal species are less able to increase their heat tolerance via acclimation than low intertidal species (Somero 2007, Berger and Emlet 2007). Thus, those species that are most tolerant to high temperatures may be the most susceptible to global temperature changes. The marine intertidal is an ideal habitat for studies of temperature-driven community change and may serve as an ‘early warning system’ for the impacts of climate warming (Barry et al. 1995, Southward et al. 1995, Helmuth et al. 2006b, Harley et al. 2006). Since organisms there live near their thermal maximum, changes in temperature are likely to elicit measurable responses. Indeed, recent episodes of warming have resulted in numerous examples of stress, mortality, and range shifts (Barry et al. 1995, Southward et al. 1995, Sagarin et al. 1999, Mieszkowska et al. 2006, Harley 2011). Research is also quite tractable in the intertidal zone because the vertical extent of the habitat is small compared to other systems (e.g., rain forests, open ocean), the body size of most organisms is small, and many species are sessile or have adult stages that are relatively easy to exclude from, or contain within treatments. Nonetheless, field studies of thermal effects on marine communities are remarkably rare given the importance of marine ecosystems. From 2000-2009, 110 experimental studies were published on the topic of ‘marine climate change’, which included all marine habitats and experiment types. Of these, 40% of studies manipulated temperature, 19% quantified impacts at the community level, and only 11% incorporated a field-based component (Wernberg et al. 2012a). Marine ecosystems provide 62% of the value of ecosystem services derived from natural

3 habitats worldwide (Costanza et al. 1997). However, climate change has and will continue to impact marine systems and their ecosystem services, with costs expected to exceed US$10 billion for some services (Gattuso et al. 2014). There is a clear necessity for marine climate change research that focuses on the higher levels of organization (i.e., communities and ecosystems) because these are the levels at which we can manage our resources. This type of research has been highlighted as an urgent research need by the IPCC; “Reliable predictions require information on multifactorial experiments performed on communities (preferably in the field), and on time scales of months to years in order to take into consideration the processes of biological acclimation and adaptation” (Gattuso et al. 2014).

1.3 Structure of this dissertation

Ecological communities of the future will undoubtedly look different than they do today. Changing global temperatures will directly influence organismal physiology, which will indirectly affect populations, species interactions, and community structure. The next step is to move beyond simple observations to a more mechanistic understanding of how warming will affect community dynamics. I approach this using manipulative field experiments to identify pathways through which temperature impacts community organization. Broadly, I ask how warming affects community assemblages via direct (e.g. organismal responses) vs. indirect (e.g. shifts in the strength or sign of interactions among species) effects. In Chapter 2, I review the literature concerning temperature and species interactions in the context of a changing climate. Despite considerable community-specific idiosyncrasy, ecological theory and a growing body of data suggest that certain general trends are emerging at local scales: positive interactions tend to become more prevalent with warming, and top trophic levels are disproportionately vulnerable. In addition, important ecological changes result when the geographic or temporal overlap between species changes. I assess the degree to which such changes are predictable, and urge advancement on several high priority questions surrounding the relationships between temperature and community ecology. In Chapter 3, I explore the population and community level patterns that result from experimentally raising environmental temperature. To do so, I created a novel in situ method for warming the substratum by approximately 2°C, using black and white settlement plates. This

4 method is used for all of my experiments. In Chapter 3, I quantify the responses of algal, barnacle, and herbivore populations at three intertidal heights over one year. I also determine how warming affected the overall community structure and diversity. In Chapter 4, I investigate one of the possible mechanisms responsible for the observed patterns in Chapter 3. Climate warming may drive organismal body temperatures beyond important physiological thresholds, directly leading to detrimental effects on populations and communities. Using two competing species of intertidal barnacles as a case study, I measure the effect of in situ warming on juvenile survivorship, growth rates, and population sizes. I describe how these similar species responded somewhat differently to increased thermal stress, due, in part, to differences in phenology. Finally, I discuss how these lethal and sub-lethal responses to warming impact adult body size and population size, which would lead to reduced fitness in a warmer world. In Chapter 5, I consider the possible indirect effects that can lead to changed community structure in a warmer world. Often, the importance of indirect effects of climate change, mediated by changing species interactions, will be greater – albeit less well understood – than direct effects in determining the community- and ecosystem-level outcomes of global climate change. To disentangle direct and indirect effects, I factorially manipulated warming and herbivore access to intertidal communities, in a 16-month long field experiment. I determined whether warming and herbivory acted independently or together to affect populations, community structure, and succession. Finally, in Chapter 6, I make some concluding remarks about work presented in the previous chapters. Specifically, I discuss the broader implications of this work and highlight future research priorities that will enhance our understanding of community change in a warmer world.

5 Chapter 2 The Influence of Temperature on Interspecific Interactions

2.1 Synopsis

Ecological patterns are determined by the interplay between abiotic factors and interactions among species. As the Earth’s climate warms, interactions such as competition, predation, and mutualism are changing due to shifts in per capita interaction strength and the relative abundance of interacting species. Changes in interspecific relationships, in turn, can drive important local- scale changes in community dynamics, biodiversity, and ecosystem functioning, and can potentially alter large-scale patterns of distribution and abundance. In many cases, the importance of indirect effects of warming, mediated by changing species interactions, will be greater – albeit less well understood – than direct effects in determining the community- and ecosystem-level outcomes of global climate change. Despite considerable community-specific idiosyncrasy, ecological theory and a growing body of data suggest that certain general trends are emerging at local scales: positive interactions tend to become more prevalent with warming, and top trophic levels are disproportionately vulnerable. In addition, important ecological changes result when the geographic overlap between species changes, and when the seasonal timing of life history events of interacting species falls into or out of synchrony. I assess the degree to which such changes are predictable, and urge advancement on several high priority questions surrounding the relationships between temperature and community ecology. An improved understanding of how assemblages of multiple, interacting species will respond to climate change is imperative if we hope to effectively prepare for and adapt to its effects.

2.2 Introduction

Temperature is one of the most fundamental determinants of biological patterns and processes. Many decades of laboratory-based research have demonstrated that variation in temperature has

6 Figure 2.1: Projected surface temperature changes for the late 21st century relative to the period 1980-1999. The panels show the multi-AOGCM average projections for the A1B SRES scenarios averaged over 2090-2099 (IPCC 2007). important and easily measured effects on biochemical and physiological rates. Because biochemical and physiological rates translate into organismal survival, growth, and reproduction, environmental temperature plays a large role in determining when and where species – particularly ectothermic species – can survive and thrive (Wethey 1983, Thomas et al. 2000, Hochachka and Somero 2002). Indeed, variation in temperature explains much of the spatial and temporal patterns we observe in the distribution and abundance of species around the world (Hutchins 1947). Although long recognized as biologically important, environmental temperature is currently being addressed with renewed vigor as anthropogenic climate change alters patterns of mean and extreme temperatures across the globe. Climate models suggest that the average temperature of the surface of the earth will warm by 1.7 - 4.4°C by the end of the current century, with increases in mean temperatures and in the frequency and magnitude of extreme temperature events (IPCC 2007). The magnitude of these projected changes varies from place to place (see Fig. 2.1). The broad-brush effects of warming are already observable across a wide variety of systems and taxa, with shifts in the distribution and abundance of species and the timing of life history events occurring largely as one would predict over spatial (e.g. latitudinal and altitudinal) and temporal (e.g., seasonal) thermal gradients (Southward et al. 1995, Sagarin et al. 1999, Parmesan and Yohe 2003, Southward et al. 2005, Helmuth et al. 2006a, Mieszkowska and Hawkins 2007). However, not every species has responded as predicted (Hawkins et al.

7 2009), and for the vast majority of species little to no data on responses to temperature exist. To better understand which species are shifting and why, and the ecological impacts of temperature changes of different magnitudes, tests of climate impacts must link processes from the climatological and biophysical to the physiological and demographic to produce a more refined understanding of how environmental temperature influences body temperature and thereby the distribution and abundance of species (Helmuth 2009). It has long been known, however, that temperature is not the sole determinant of where a species can live and how well it will perform. For example, Darwin (1859) recognized that many distributional patterns across thermal gradients seemed to depend more on interactions among species than upon the direct effects of temperature, an observation that has since received extensive observational and experimental support (Connell 1961a, MacArthur 1972). The current theory holds that a species’ response to spatial or temporal variation in temperature will depend both on direct effects on the individual- and population-level attributes of that species and on indirect effects mediated by changes in the distribution, abundance, and behavior of competitors, predators, parasites, and mutualists (Dunson and Travis 1991, Davis et al. 1998, Sanford 1999, Hawkins et al. 2009, Wernberg et al. 2011, Johnson et al. 2011). Thus, although general patterns of change may be robust and predictable (Barry et al. 1995, Parmesan and Yohe 2003), accurate predictions regarding the consequences of warming for particular species or ecosystems of interest often remain elusive. A significant challenge in this era of global change is to improve our predictive power with regards to the ecologically important consequences of climatic warming. To accomplish this, we must integrate single species, ecophysiological / population-level approaches and multispecies, community- and ecosystem-level research into a single framework so that general hypotheses regarding the effects of warming can be formulated and tested, and a theory of climate change ecology can progress. Here, I consider biological effects of temperature change across levels of organization from enzymes to ecosystems to determine how much is known about the potential effects of temperature on complex groups of interacting species. I begin with a brief review of how temperature affects basic metabolic processes, and then explore how differences in these responses among species affect species interactions. Next, I consider how differences in physiological responses across different species can influence the overall effect of temperature on ecological communities. Finally, I outline possible frameworks for generalization

8 of the impacts of temperature on ecological systems, and consider broader implications of these generalities for climate change and biogeographic patterns in marine systems. I do not intend to present an exhaustive review of the ever-expanding literature on climate change. Rather, I aim to highlight the ways in which warming will influence species interactions, and the ways in which species interactions will determine the outcome of warming.

2.3 The biological importance of temperature

Temperature is one of the most important factors affecting biological processes in poikilotherms (see Fig. 2.2 for a summary). The link between temperature and biological processes is kinetic; as temperature rises and atoms become more energetic, processes such as diffusion speed up and molecules in a fluid collide with one another more frequently. For enzyme-catalyzed reactions, higher temperatures increase the likelihood that enzymes will collide and bind with substrate molecules during a given time frame, enhancing the speed and efficiency of biochemical reactions. However, enzymes are proteins that are largely held together by hydrogen bonds, and temperatures that exceed some threshold can weaken these bonds, causing proteins to change shape and thus reducing or negating their effectiveness as biological catalysts. Because enzymes work best within a specific temperature range, and because diffusion increases with temperature, catalytic rates typically increase with temperature to a point after which they fall off rapidly (Campbell and Farrell 2006) (Fig 2.3a).

9

Figure 2.2: The pathway by which temperature as a physical phenomenon influences the ecology of individuals and populations.

10 100 (a) 75

50

25

Enzyme activity (%) 0 0 10 20 30 40 50 60 10 ) (b) -1 8 min

-1 6 kg

2 4 2 Scope for work (mg O 0 8 10 12 14 16 18 20 0.55 (c) )

-1 0.5

0.45 (mm day

Individual growth rate 0.4 20 25 30 35 2 (d) ) -1 1 (day

0 Population growth rate 0 5 10 15 20 25 30 Temperature (°C)

Figure 2.3: Relationship between temperature and various biological rates for representative species (note the differences in x-axis scale). (a) Activity of the enzyme lactate dehydrogenase in the fish Champsocephalus gunnari (Coquelle et al. 2007). (b) Metabolic scope for work (measured as maximal metabolic rate minus resting metabolic rate) in sockeye salmon Oncorhynchus nerka (Lee et al. 2003). (c) Individual growth rate in the Cortez oyster Crassostrea corteziensis (Cáceres-Puig and Abasolo-Pacheco 2007). (d) Population growth rate of the marine diatom Phaeodactylum tricornutum (Kudo et al. 2000), using data for iron-replete cultures.

11

Enzymatic reactions underlie functions at higher levels of organization; therefore, other biological rates often exhibit similar relationships with temperature. For example, metabolic function is strongly temperature dependent. For ectotherms, rising temperature increases the rates of basal metabolism and the rate at which energy stores are depleted. Temperature also determines the maximum metabolic rate, which determines the limits of non-maintenance activities such as exercise (via the breakdown of energy stores) and growth and reproductive investment (via the build-up of somatic and gonadal tissue). The difference between the active metabolic rate (the maximum rate at which an organism can expend energy, e.g., during activity) and the resting metabolic rate (the rate at which an organism must expend energy to stay alive and healthy, e.g., respiration) can be thought of as the metabolic scope for work. In essence, the metabolic scope for work is a proxy for the energy available for non-maintenance functions such as physical activity, growth, and reproduction (metabolic scope for work is therefore a broader term than the more commonly used ‘metabolic scope for activity’; e.g. Claireaux and Lefrançois 2007). As with biochemical reactions, scope for work increases from low temperature towards some optimum, and then begins to fall off as costs begin to accrue more rapidly than benefits (Lee et al. 2003) (Fig. 2.3b). Because metabolic scope for activity represents energy available for non-maintenance functions, it is not surprising that individual growth rates display a similar unimodal relationship with temperature (Fig. 2.3c). Note that the temperature-growth relationship depends on food and other resources being amply supplied; if food is scarce, an organism may not meet its maintenance metabolic costs even though it is capable of high levels of activity. I will return to this idea when I discuss the Metabolic Theory of Ecology. Faster individual growth rates in turn tend to reduce generation time, and thermal control of generation time has important consequences for rates of population growth (Huey and Berrigan 2001). Indeed, population growth rates measured in the lab frequently exhibit the same relationship with temperature as individual growth rates (Fig. 2.3d).

12 2.4 Interspecific variation in thermal sensitivity

Every species will exhibit some relationship between temperature and fundamental biological performance parameters such as metabolic rate and growth. However, the relationship between temperature and performance can vary widely among species. There are two fundamental ways in which this interspecific variation can manifest: 1) differences in thermal sensitivity (i.e., the slope of the temperature:performance relationship), and 2) differences in the maximum, minimum, or optimal temperatures for a given biological function. Variation in thermal sensitivity is diagrammed in Fig. 2.4a. In my hypothetical example, one species exhibits a relatively large increase in performance from low to optimal temperatures (Fig. 2.4a, dashed line), while another has a much more gradual increase in performance over the same range (Fig. 2.4a, solid line). Although both species have the same thermal range, the latter species (solid line) is less sensitive to changes in temperature, and will outperform the more thermally sensitive species at colder temperatures (point x) but not at warmer temperatures (point y). Alternatively, interacting species can have a difference in the position of the peak of their performance- temperature curve and in their thermal limits (Fig. 2.4b). In this case, the species represented by the solid line outperforms the other species at low temperature (point x), but is lost from the system at higher temperatures (point y).

13

Figure 2.4: Interspecific variation in the impact of rising temperatures. In the upper panel (a), the species represented by the dashed line is more sensitive to changes in the thermal environment across most temperatures, but each species has the same thermal range. In the lower panel (b), the dashed-line species has a higher upper thermal limit. If ‘ecological performance’ were to represent, e.g., competitive ability, an increase in temperature from x to y would result in a shift in competitive dominance from the solid-line species to the dashed-line species.

Interspecific variation in thermal sensitivity is a general phenomenon. Before I present illustrative examples, however, we need a metric to describe the relationship between temperature and performance so that we may more easily compare thermal sensitivity among species. One such metric is the Q10 value, which is the factor by which performance (e.g., enzymatic reactions, metabolic rate, growth) increases with a 10°C increase in temperature. For example, a Q10 of 2 implies a doubling of metabolic rate when temperature is increased from

10°C to 20°C (Gillooly et al. 2002). When Q10 values are compared among interacting taxa, they may vary considerably; Q10 values for northern European bivalve metabolic rates are near 2.0, while the Q10 values for the metabolic rates of species which prey on those bivalves can range from 1.5 to 2.5 (Freitas et al. 2007). When two species with different thermal sensitivities are allowed to interact, the outcome of that interaction is also temperature sensitive. For example, predatory flagellates are more sensitive to (i.e., respond more positively to) increases in

14 temperature than do their bacterial prey (Delaney 2003). Although Delaney (2003) was primarily concerned with the effects of turbulence, approximate Q10 values can be calculated from the data presented in her Tables 1 and 2 (using the turbulent treatment, which was considered a better approximation of natural conditions). The Q10 for the population growth rate of the predator (~3.4) was higher than that of the prey (~2.4), which would correspond to the dashed and solid lines in Fig. 4a, respectively. As a result of both relatively more rapid predator population increases and higher per capita predator ingestion rates at higher temperatures, the overall mortality of bacteria due to flagellate grazing increased over 5-fold for every 10°C of warming (Delaney 2003). Rising temperatures could therefore favor bacterial population growth in the absence of a predator but hinder bacterial population growth in the presence of a predator. Not surprisingly, variation in thermal range or thermal optima among species within a community is also a widespread phenomenon that has important ecological consequences. For example, on New England rocky shores, two competing species of barnacles have different maximum temperature tolerances, and a combination of temperature and interspecific competition determines the distribution of the two species. In cooler, northern areas, thermally intolerant Semibalanus balanoides (represented by the solid line in Fig. 2.4b) competitively excludes the more thermally tolerant Chthamalus fragilis in the mid and high intertidal zones (dashed line, Fig. 2.4b). In warmer southern areas, high temperatures exclude S. balanoides from the higher shore levels and C. fragilis occupies that free space (Wethey 1983, 1984a). Similar relationships occur on European rocky shores with S. balanoides outcompeting Chthlamalus species, except at very close levels (Connell 1961a). The importance of climatic fluctuations in mediating interactions between S. balanoides and Chthlamalus species, have been long known (Southward and Crisp 1954, Southward 1991). Recent analysis of 40 year data sets and modeling (Poloczanska et al. 2008) have shown that Chthamalus species are released from competition with the faster growing, cold-water S. balanoides in warm years. As illustrated by the above examples, warming temperatures can affect a species via both direct and indirect pathways. There has been a great deal of emphasis on the direct impacts of temperature on ecological variables including local abundance. However, indirect effects such as the increase in C. fragilis observed when high temperature inhibits the dominant competitor may also be just as important (Poloczanska et al. 2008). These indirect effects can be divided into two categories: per capita effects, where temperature changes the strength of a single individual’s

15 interaction within a community, and density effects, where temperature changes in the total number of individuals in the population. Both mechanisms can and probably do operate simultaneously. For example, during periods of upwelling relaxation, when sea surface temperatures increase, the sea star Pisaster ochraceus (Fig. 2.5a) becomes more abundant in the intertidal zone where it forages due to increased mobility (a population-level effect) (Fig. 2.5b). In addition, individual Pisaster consumes more mussels per unit time in warmer water (a per capita effect) (Fig. 2.5c). The net effect of warmer water is a dramatic increase in the rate of mussel mortality due to predation (Sanford 1999). Although mussels do grow more quickly in warmer water (Menge et al. 2008), a warming-induced increase in predation may more than offset this direct positive effect on mussel populations. Indirect effects mediated by species such as Pisaster may determine much of the net effect of warming at the community and ecosystem levels. As noted by Sanford (1999), key species interactions that are sensitive to temperature may act as “leverage points” through which small changes in climate could generate large changes in natural communities. Species that act on these leverage points can amplify the signal of small changes in climate to generate unexpectedly large changes at the community level. In addition to classic keystone species such as Pisaster, many diseases and pests are likely to operate on leverage points. For example, warming increases the incidence and impact of pathogens in many marine species (Harvell et al. 2002), including Pisaster (Bates et al. 2009). This further highlights some of the potential complexities involved; Pisaster predation may increase with temperature, but over the longer term this effect may depend on the presence and epidemiology of sea star disease agents.

16 Figure 2.5: Direct and indirect effects of rising temperature (T) on an interacting species pair. (a) Pisaster ochraceus and Mytilus californianus. (b) Density effects, where rising temperature increases Pisaster abundance (solid, thick red arrow). (c) Per capita effects, where the strength of predation (black arrow) is increased (made more negative) by rising temperatures (solid, thick red arrow). In both the per capita and density-mediated cases, the net effect of temperature on mussels is negative (dashed red line) despite any weak direct effects to the contrary (thin red line in panel b).

2.5 Incorporating time and space: phenology and biogeography

A change in temperature can alter species interactions if the sign or magnitude of response differs among the species (Fig. 2.4). Species interactions may also change if temperature causes a change in the temporal or spatial abundance pattern of one of the species relative to another. Climatic warming is causing spring to start earlier and summer to last longer (Menzel and Fabian 1999, Thompson and Clark 2008), and as a result many plant and animal phenologies (the timing of reproduction, larval release or settlement, fledging, migration, etc.) are also shifting earlier (Sims et al. 2001, Philippart et al. 2003, Edwards and Richardson 2004, Hays et al. 2005). Parmesan & Yohe (2003) showed that over 45 (median) years 62% of 678 species worldwide have exhibited changed phenologies. In addition, a meta-analysis of 203 species spanning the northern hemisphere revealed an advance in spring-cued phenology of 2.8 days/decade (Parmesan 2007), and coastal marine species are moving even faster (Helmuth et al. 2006b). Moore et al. (2011) have recently showed that whilst a southern species of limpet (P. depressa) is breeding earlier and longer, a northern autumn breeding congener is breeding later and failing to breed in some years. The timing of life cycle transitions must often be in (or out of) synchrony with the phenology of other species, particularly when those species represent an important food resource

17 or an important source of mortality. For example, the timing of hatching or spawning often occurs when food resources will be most plentiful for offspring (Platt et al. 2003). Mismatches between periods of larval presence and planktonic food abundance associated with interannual climate variability have been long been blamed for poor fisheries yields (Cushing 1982). Although consumers can be cued by their resource directly, many must rely on some perceptible environmental cue such as temperature or light as a proxy for it. Although most documented cases of this phenomenon have been from terrestrial systems, recent work in marine systems, primarily on seabirds and pelagic communities, have highlighted how linked species can be cued by different factors (Costello et al. 2006, Richardson 2008, Watanuki et al. 2009). For example, Edwards and Richardson (2004) analyzed data from 66 marine taxa spanning more than 40 years and found that diatom blooms have remained fixed in time (cued by light) while temperature- cued consumers have shifted reproduction earlier as summer water temperatures increase. This has led to a phenological mismatch between trophic levels. Differential use of the thermal landscape can also lead to temporal mismatches. Thermal cues in migratory animals’ wintering grounds are becoming less predictive of conditions on the breeding grounds. Indeed, many migrant animals rely on a series of locations during the year, each with a different climatic regime, each changing at a different rate with global warming (Jonsson and Jonsson 2009). Historically, migrant animals have arrived at their breeding grounds in synchrony with their food source. However some but not necessarily all sites along a migration route are being affected by warming, thus animals end up mistimed with their resource at their reproductive locations (Carscadden et al. 1997, Sims et al. 2004). As formerly relevant seasonal cues lose their accuracy in matching resources and environmental conditions, phenological mismatches are becoming common. One study reviewed cases where species had become mistimed to see if they had fallen too far out of alignment, and found that out of 11 cases, eight had become uncoupled, shifting either too soon or too late compared to the other (Visser and Both 2005). The most pertinent question may be whether these mismatched species remain uncoupled, or whether ecological or evolutionary processes can compensate for negative consequences of the mismatch. For example, selection or plasticity in phenology could act strongly enough to re-couple them over time, or to facilitate prey switching or other behavioral shifts to compensate for climate impacts.

18 Biogeographic range shifts are another obvious biological manifestation of climatic warming (Helmuth et al. 2006b); 75% of 129 coastal marine species have undergone poleward shifts in their geographic distributions, at an average rate of 19km/year (Sorte et al. 2010). Warming-induced range shifts may widely alter the complement of interacting species at a site (Cheung et al. 2009), and interspecific interactions may determine the extent to which any given species range changes with warming. I consider each of these scenarios in turn. Biogeographic range shifts during times of environmental change are nothing new in the earth’s history, and the fossil record can shed a great deal of light on the implications of ongoing and future range shifts. Analyses of post ice age warming in the late-Quaternary indicates that some species shift their range limits during periods of warming while others do not (Roy et al. 2001). Such individualistic responses among species can cause historically separated species ranges to converge, potentially generating a new interspecific interaction, or force interacting species apart geographically and eliminate an interspecific interaction (Fig. 2.6). This reshuffling of taxa results in combinations of species that cannot be found together anywhere on earth at present – a situation known as a no-analog community (Williams and Jackson 2007). There are several marine examples of no-analog communities during the recent geological past when global temperatures differed considerably from the present (Kitamura 2004, Steinke et al. 2008), and more no-analog communities can be expected in the future. One critical question is whether no-analog communities differ in their structure or functioning relative to communities in which evolution may have led to sets of traits that allow greater function or unique community structure. Many no-analog communities already exist as a consequence of human mediated species introductions. While there is substantial evidence that biological invasions can change community structure through cascades of interactions (Grosholz et al. 2000, Wonham et al. 2005), there is not clear evidence that novel combinations of species within a community consistently alter community structure or functioning relative to uninvaded communities, though the role of temperature in this context has not been examined explicitly.

19

Figure 2.6: Hypothetical range shifts due to global warming with the resulting species interactions. (a) The yellow species’ historical range overlaps with that of the red species. (b) Warming may cause species’ ranges to move poleward, but to different extents, generating novel interactions, such as with the yellow and blue species.

There is no doubt that biogeographic changes like those recorded in the fossil record are ongoing today. In recent decades, warming has triggered an expansion of species’ poleward range boundaries and a contraction of equatorward range boundaries (Southward et al. 1995, Sagarin et al. 1999, Perry et al. 2005, Southward et al. 2005, Helmuth et al. 2006b, Moore et al. 2007b, Sorte et al. 2010). The velocity of these modern shifts can be striking; the southern range limit of a barnacle and the northern range limit of a benthic polychaete are moving north at rates of 15-50 km/decade in Europe (Wethey and Woodin 2008), and some planktonic species are moving an order of magnitude faster than that (Beaugrand et al. 2002, Hays et al. 2005). While some species ranges are shifting quickly, others are shifting slowly, and still others are either not shifting at all or are moving in the opposite direction (Perry et al. 2005, Lima et al. 2007). As with no-analog communities of the past, this complex redistribution of species guarantees that some species will exchange encounters with familiar organisms for interactions with novel organisms. For example, global warming is facilitating the poleward spread of many harmful algal bloom species, creating risks for wildlife and human health in previously unimpacted areas in both hemispheres (Hallegraeff 2010).

20 Changes analogous to these latitudinal shifts are also occurring across vertical gradients of depth and intertidal height. The depth distributions of North Sea fishes are generally shifting to deeper waters, although the degree and even direction of depth range change is species- specific (Perry et al. 2005). Although the ecological implications of any resulting shifts in interspecific interactions remain largely unknown for fish assemblages, some data are available for redistributions of benthic species across the vertical gradient. On rocky shores in the northeast Pacific, rising temperatures have forced the upper limits of the alga Mazzaella parksii to lower positions on the shore (Harley and Paine 2009). Although the upper limit of the alga is related directly to temperature via the species’ environmental tolerance, the lower limit (set by molluscan grazers) is independent of temperature. Higher temperatures result in an increase of the spatial overlap between the potential vertical range of Mazzaella and that of its consumers, which in turn leads to the elimination of the alga in warm areas which lack a spatial refuge from herbivory (Harley 2003). This latter example is illustrative of the potential role of interspecific interactions in determining the degree to which species ranges may expand or contract with warming. To remain within its current thermal envelope, Mazzaella would have had to shift both its upper and lower limits downshore. However, consumers prevented such a shift in the lower limit, with negative implications for the total vertical range of the alga. Interactions among species are known to determine the position of range limits along a thermal gradient in the laboratory (Davis et al. 1998). The degree to which species interactions may generally facilitate or inhibit a species’ ability to track its preferred environmental conditions (often called its bioclimatic envelope) in the field, particularly at larger spatial scales, remains an open question.

2.6 The search for generality

Although the effect of temperature on the performance of an individual, population or species varies from case to case, considering generalities of biological effects of temperature allows the articulation of testable hypotheses and exploration of potentially broad-scale impacts of temperature on communities and ecosystems. Recent work suggests that interspecific interactions shift from generally negative (e.g. competitive) when the environment is benign to generally positive (e.g., facilitative) when the environment is stressful (Bruno et al. 2003). Since

21 high temperature can qualify as an environmental stress, many interactions are predicted to shift from competitive to facilitative at higher temperatures (Wernberg et al. 2010). This has been shown to occur within a community-type; for example, canopy-forming algae on rocky shores compete with barnacles for space at cool sites but facilitate them by providing cool understory microhabitats at warm sites (Leonard 2000). Facilitation theory is relevant to systems where species interactions can ameliorate physical or physiological stress, and intertidal rocky shores or marshes are emblematic habitat types for facilitation. It is less clear how the prevalence, strength or importance of facilitation will change in subtidal communities where organisms cannot modify the temperature of the ocean. However, this difference in facilitation across stress gradients is contradicted by a comparison among community types (e.g. warmer high-shore barnacle-dominated communities vs. cooler low-shore kelp-dominated communities) in which the relative importance of competitive and facilitative interactions does not appear to change (Wood et al. 2010). Facilitation may also favour one species more than others: Moore et al (2007b) showed that the behaviour of the Northern species of limpet Patella vulgata allowed it to benefit from habitual amelioration by fucoid clumps; whilst its more southerly congener, P. depressa did not display such behaviour. These changes have implications for patch dynamics and functioning of European rocky shores (Hawkins et al. 2008, 2009). The broader search for generalities in ecology has led to the development of the Metabolic Theory of Ecology (MTE), which relates metabolic rate to body size and temperature (Gillooly et al. 2001). MTE predicts that metabolic rate increases with temperature in specific ways across broad taxonomic groups (Gillooly et al. 2001). However, at this coarse resolution, differences among some groups persist. Exploring these differences at the group level (e.g., primary- versus secondary-producers, fish versus invertebrates (Gillooly et al. 2001, Lopez- Urrutia et al. 2006)) may lead to general patterns in how community structure (relative abundance of species or functional groups) varies with temperature change. In this way, a theory of how temperature affects community structure can be developed and tested. Specific physiological rates may also respond differently to changes in temperature. For example, both theoretical and empirical evidence suggests that marine planktonic respiration increases more rapidly with rising temperature than does photosynthesis (Lopez-Urrutia et al. 2006). Thus, rising temperatures should shift marine planktonic systems away from autotrophy and towards heterotrophy, a prediction which has some empirical support (Müren et al. 2005).

22 MTE also makes predictions regarding the relationship between temperature and food web structure. Food chain length depends on the amount of energy transferred through trophic interactions. For a given (fixed) resource base, the highest trophic level in the system is that which can support a minimum viable population with the energy available from lower trophic levels (Oksanen et al. 1981). As temperature increases, the metabolic rates of all species increase, resulting in an increasing demand for and consumption of energy at each trophic level. When the supply of energy transferred up the food chain is no longer sufficient to support the minimum viable population size of the top predator, that species is lost. There is some empirical support for this prediction; by experimentally warming mesocosms containing aquatic microbes, Petchey et al. (1999) found that higher trophic levels were lost disproportionately, and food chain length decreased. As consumers were lost in warmed treatments, primary producer and bacterivore biomass increased; suggesting that a thermally-triggered trophic cascade had occurred (Petchey et al. 1999). Although these results are consistent with MTE predictions, other alternatives, such as lower physiological tolerance to warming in species at higher trophic levels, cannot be ruled out. Furthermore, in surface and coastal marine systems, ocean currents and nutrient availability change with temperature in varied ways. Changes in upwelling will increase nutrient availability while constraining temperature changes, while in other areas increased thermal stratification will reduce nutrients concurrent with warming. Changes in upwelling will also influence recruitment regimes (Menge et al. 2011). Any general effects of temperature on species interactions will occur in the context of other, potentially more influential environmental changes. The principal challenge at this stage is to develop and test predictions for how these changes interact to influence species interactions to determine whether any generalities exist. MTE has been useful for generating broad-scale models of the ecological responses to temperature change. It is less clear whether MTE applies to smaller spatial and temporal scales, where species' traits and differences may be more important. In this case, more detailed theories like dynamic energy budgets may be more relevant for generating predictions (Helmuth et al. 2006a).

23 2.7 Future research priorities

Anthropogenic climate change is creating an ongoing series of challenges for human societies that rely on natural goods and services. At present, our lack of understanding of the interplay between temperature and interspecific interactions prevents ecologists from making anything more than relatively basic predictions regarding the effects of warming on community structure, on ecosystem function, and even on individual species of concern. The degree to which future outcomes will follow predictable patterns based on general species attributes (e.g., trophic level) or will only be predictable with careful study of the individual species involved remains unclear. In either case, predictions for the future inherently require extrapolation beyond the current range of observations, and therefore require the application of basic, mechanistic ecological principles to new situations (Poloczanska et al. 2008). A stronger mechanistic understanding of climate change impacts can be achieved through a systematic approach that emphasizes the testing of hypotheses in experimental frameworks (Firth et al. 2009, O'Connor et al. 2011). This fundamental scientific method has not been emphasized in climate change ecology, in part because the focus has been on documentation of impacts. The current challenge is now to determine the extent to which we can understand the causes and consequences of these impacts in a general ecological framework. Currently, numerous hypotheses based on physiological, ecological and evolutionary theory can be articulated and experimentally tested. I outline a few key questions here: • Can among-species variation in thermal sensitivity (i.e., the slope of the temperature: performance relationship) or critical temperatures (thermal optima, maximum or minimum temperatures for a given biological function) predict how interactions such as competition and predation will change with warming? Some evidence suggests that this approach may bear fruit, particularly for trophic relationships where production and consumption rates can be carefully measured (Delaney 2003). However, at least one classic competition example (Park 1954) shows that surpassing the growth rate of a superior competitor at higher temperature does not lead to a switch in competitive dominance, and simple comparisons of growth rates may be misleading in light of the potential trade-off between growth rate and competitive ability.

24 To what extent will ecological change be driven by changes in abundance of interacting species (population-level effects) vs. changes in per capita effects? Although much of the ecological literature focuses on the relative change in abundance of strong interactors (e.g., predators, ecosystem engineers, disease vectors), which is easier to measure, the flour beetle example mentioned above (Park 1954) along with more recent work (Sanford 1999, Moore et al. 2007a) suggests that per capita interactions may be critical. Integrating these two levels of impact is a priority because they can driven by different mechanisms and therefore may change at different rates with environmental change, and be subject to different constraints and limitations. • Is the shift from predominantly negative interactions to predominantly positive interactions as stress increases – a phenomenon which holds for specific, defined assemblages (Leonard 2000) in habitats where organisms can modify the thermal environment – likely to apply when species composition is also changing? What is the role of facilitation in subtidal systems where organisms are not able to modify their thermal environment? • How much can the Metabolic Theory of Ecology tell us about specific communities? Is the loss of top predators during periods of warming a general phenomenon? And, as with the question of positive versus negative interactions, does the decrease in food chain length only apply when novel, thermally tolerant species are not allowed to invade the system? • To what extent will evolution minimize or even exacerbate community-level responses to warming? Local adaptation to the thermal environment is well documented, and mechanistic predictions developed using present-day thermal tolerance limits, temperature-performance functions, or phenological relationships to temperature may not apply in the future.

Answers to these questions will require studies that simultaneously address physiological responses to abiotic variables and ecological relationships among interacting species. Dunson and Travis (1991) lamented the scarcity of such studies almost two decades ago, and there is still a great need to unify ecophysiology and community ecology. Such research will be necessary to field-test hypotheses that have been developed on the basis of thermodynamic considerations and laboratory results. Studying ecological dynamics in artificially warmed areas such as power plant

25 cooling water discharge plumes (Schiel et al. 2004) or during warm phases of natural climatic cycles (e.g. ENSO) is a good start, but well-designed thermal manipulations (Harte and Shaw 1995, McKee et al. 2003) that test responses of critical ecological and evolutionary processes in the context of theory are badly needed. Furthermore, although much can be learned from the paleo-ecological perspective, ongoing research must incorporate potential synergisms between warming and other modern anthropogenic effects such as habitat modification, species introductions, over-exploitation, pollution, and elevated carbon dioxide (Williams and Jackson 2007). Finally, our current predictions of future change are founded on present-day physiological and ecological responses to temperature, but organisms can acclimate and populations can evolve. Although there has not yet been any evidence of genetic changes in populations towards higher thermal tolerances, populations can track climatic shifts to varying degrees through genetic change or plasticity (Bradshaw and Holzapfel 2006). The extent to which phenotypic plasticity and natural selection will offset the effects of warming (Jones et al. 2008) is poorly understood at best. Although the challenges are many, a fuller understanding of the complexities surrounding community-level responses to warming is a prerequisite for successfully predicting, mitigating, and managing the effects of global warming.

26 Chapter 3 Intertidal Community Responses to Field- Based Warming

3.1 Synopsis

As the climate warms, there is little doubt that ecosystems of the future will look different from those we see today. However, community responses to warming in the field are poorly understood. I examined the effects of field-based warming on intertidal communities in the Salish Sea, which is a regional thermal ‘hot spot’ and therefore a model system for studying thermally stressed communities. I manipulated temperature at three tidal heights by deploying black- and white-bordered settlement plates. Black plates increased in situ substratum temperature by approximately 2°C. Barnacles fared poorly on black plates in all zones. When overall thermal stress was highest (summer in the high intertidal zone) herbivores were absent. In lower tidal zones, herbivores were abundant on white plates but were scarce on black plates. The total percent cover of algae was unaffected by the temperature treatment, despite the fact that macroalgae were expected to be the least thermally tolerant functional group. However, I did find that ephemeral green algae exhibited a delay in phenology on black plates. I also found that species richness declined and invertebrate assemblage structure was altered due to warming. Results from this year long experiment suggest that communities in thermally stressful habitats respond to warming via the interplay between species-specific thermal responses and secondary adaptive strategies such as behavioral microhabitat selection. Declines in diversity and changes in the invertebrate assemblage were due to the decline of local thermally-stressed species and the lack of replacement by warm-adapted species. Given the low variation in the species pool along the northeast Pacific coastline, the arrival of warm-adapted species to the Salish Sea may not occur over relevant time scales, leaving local communities depauperate.

3.2 Introduction

The earth’s climate is changing, with models predicting an average 1.0-3.7°C rise in global surface temperature by the end of the century (IPCC 2013). This warming will take the form of

27 both increasing mean temperatures and increasing frequency and severity of thermal extremes (Trenberth 2012, IPCC 2013). Ongoing consequences of anthropogenic warming have been ubiquitously observed in a variety of systems and taxa (Parmesan and Yohe 2003, Pinsky et al. 2013). Themes are emerging at the species and population level, including poleward shifts in species distributions and changes in the seasonal timing of life history stages, and demographic shifts towards smaller sized individuals (Parmesan and Yohe 2003). However, much is still unknown about the responses of whole communities given the expected reshuffling of assemblages and changing interactions (Lurgi et al. 2012). Our ability to predict responses to altered climatic patterns relies on a coherent understanding of how warming affects all levels of organization; from individual- to biome-level processes. Community level predictions are particularly challenging due to individualistic species responses to abiotic and biotic changes. In addition, species respond to temperature differently when interacting with other species (Davis et al. 2002). For example, abiotic stress may remove otherwise dominant competitors, allowing subordinate species to persist (Poloczanska et al. 2008) or shift negative relationships to positive ones (e.g. Leonard 2000). Because the direct effects of climatic warming can vary in sign and magnitude among interacting taxa, indirect effects mediated by shifts in interspecific relationships may be common. Resolving the ways in which the myriad of idiosyncratic species responses will drive these indirect effects and changes in community-level patterns remains one of the greatest challenges in climate change ecology. One approach is to quantify species responses to experimental warming within a community framework, in situ. Since direct and indirect species interactions can modify how taxa respond to warming (Davis et al. 2002), multi-species studies can provide a more accurate and holistic representation of community responses because they capture direct and indirect responses. Further, more accurate climate predictions can be made by relying on a combination of observational studies, lab and mesocosm experiments, and in situ field experiments (Wernberg et al. 2012a, Stewart et al. 2013). Field manipulations can mimic projected abiotic stress, have appropriate controls, and a realistic species pool (a combination of qualities not possible in other experimental methods), providing a crucial experimental link between observational and laboratory studies. Nonetheless, field experiments are comparatively lacking in climate change ecology (except in terrestrial plant communities; Wolkovich et al. 2012).

28 The scarcity of field experiments is particularly noticeable in marine environments, where documentation of climate change impacts lags behind that of terrestrial environments (Richardson and Poloczanska 2008). The ratio of experimental to observational studies is quite low due to the logistical difficulties inherent in manipulating temperature in marine systems (Wernberg et al. 2012a). This is due in part to the thermal properties of water; it is much more difficult to manipulate body temperature in water than air, particularly in the field where it is generally impractical to heat or cool the large volumes of water that move past an organism and largely determine its body temperature. Nevertheless, a limited number of opportunistic studies have shown that community composition changes with warming in the field (Schiel et al. 2004, Wernberg et al. 2012b). The importance of temperature is further supported by numerous marine laboratory and mesocosm experiments, the vast majority of which demonstrate significant effects of warming (reviewed by Wernberg et al. 2012a). Intertidal systems have become a popular laboratory for climate change ecology, and represent a tractable system for in situ warming studies. Some key organisms in the intertidal already live at or near their thermal limits (Hofmann and Somero 1995, Davenport and Davenport 2005) and recent episodes of warming have resulted in stress, mortality, and range shifts (Southward et al. 1995, Harley 2011). Thus, the intertidal may serve as ‘an early warning system’ for the impacts of climate warming (Helmuth et al. 2006b). Intertidal organisms are unique because they exist in both terrestrial and marine environments; the former being more variable and the latter being more homogeneous in temperature. In addition, the short generation time and rapid turnover rates of many intertidal species allows researchers to detect responses to field manipulations relatively quickly. Although laboratory-based estimates of thermal tolerance are available for several intertidal taxa, allowing for some basic predictions to be made, the effect of intertidal warming at the community level is poorly understood. In this study, I used passively warmed settlement plates in situ to test the effects of warming on intertidal communities. I hypothesized that aerial warming during low tide would be detrimental to intertidal species because some already reach stressful body temperatures when exposed (Hofmann and Somero 1995). Specifically, I expected that, on average, prostrate macroalgae would fare poorly due to thermal stress. I expected that mobile grazers would be less abundant with localized warming because their food source would be limiting or because they would avoid warmed treatments. If grazers avoided warm treatments, then palatable algae would

29 be released from herbivory, increasing their abundance. I also expected that thermal stress would reduce barnacle populations, however, I expected the effect to be milder compared to other taxa, due to the ability of barnacles to withstand more severe thermal stress (as evidenced by their ability to live higher on shore). Overall, I predicted that warming would reduce species richness by eliminating thermally sensitive species and shift species composition away from food webs based on in situ benthic production (algae, limpets) and towards thermally tolerant species that do not rely on local benthic primary productivity (e.g., barnacles).

3.3 Methods

3.3.1 Study system

This experiment was conducted on the southeast shore of Salt Spring Island, one of the southern Gulf Islands in British Columbia, Canada (48.753° N, -123.388° W). Tides around the Gulf Islands are mixed semidiurnal with a maximum tidal amplitude of 3.9 m during the experimental period. Daytime lower low tides occur only during summer months (March-August) and coincide with the hottest portion of the day (10am-3pm). In addition, the Gulf Islands are protected from both oceanic swell and wind waves by Vancouver Island (a large barrier island directly west of the island chain; Demes et al. 2012), and fall within the rain shadow of Vancouver Island mountains. The archipelago that includes the Gulf Islands (Canada) and San Juan Islands (USA) is considered a ‘hot spot’ for stressful intertidal temperatures relative to the outer west coast (Helmuth et al. 2002). These conditions make Salt Spring Island an ideal location for the study of thermal stress on intertidal communities.

3.3.2 Experimental warming treatment

During sunny days with calm low tides, organismal body temperatures can exceed air temperatures by 10-15°C (e.g. Denny and Harley 2006). Thus, body temperature provides a more useful proxy for thermal stress than air temperature (Helmuth et al. 2005). For many intertidal organisms, body temperature is tightly tied to the temperature of the substratum because the organismal heat budget is dominated by conduction rather than convection (Denny and Harley 2006). I took advantage of this to raise body temperatures in situ by warming the substratum and

30 thus increasing the amount of heat transferred to an individual via conduction. This method is most appropriate for organisms with a high surface area in contact with the substratum, such as limpets and barnacles (Denny and Harley 2006). In addition, body temperatures of littorine snails more closely match substratum than air temperatures, but can exceed both (McMahon 1990). Experimental heating via conduction is less useful for erect algae such as Mastocarpus papillatus because convection and evaporation are more important to its heat budget (Bell 1995), although tropical seaweeds have surface body temperatures within 0.5-1.5°C of substratum temperatures (Cox and Smith 2011). Conductive heating is likely most important for prostrate algae such as crusts, diatoms, and green filamentous algae during emersion, which have a high surface area in contact with the rock. I used black and white bordered settlement plates that would passively heat the substratum without changing irradiance (e.g., by shading) to alter the temperature of a surface conductively connected to intertidal organisms. Due to differences in absorption, the warmed (black) treatment attained higher temperatures during daytime low tides. I installed black and white settlement plates at each of three tidal heights (1.1 m, 1.6 m, 2.6 m above chart datum) in the rocky intertidal. The Canadian chart datum is calculated using the lowest astronomical (uninfluenced by weather) tide approximated by the Canadian Hydrographic Service. Settlement plates (Fig. 3.1) were made of black or white 0.56 cm High-Density Polyethylene (HDPE) “puckboard”. Each HDPE plate was 15.25 x 15.25 cm with a centered 6.9 x 6.9 cm area of white textured epoxy (Sea Goin’® Poxy Putty HD; Permalite Plastics) 3 mm thick above the HDPE. Unlike previous settlement plate designs, I used the same material on all plates for the central settlement area because differences in chemical composition and substratum color influence recruitment independent of thermal differences (Swain et al. 2006, Finlay et al. 2008). Several holes were drilled in the center of the HDPE plate and the surface was gouged to promote epoxy adherence to the surface, as it otherwise tended to peel up after months of field use. Previous studies have shown that textured detoxified epoxy putty is a good settlement surface for algal zygotes as well as invertebrates such as barnacles (Kordas and Dudgeon 2011). To create a textured settlement surface that was uniform across plates, I pressed a single layer of salt (sieved to 1-2 mm crystal size) into the surface of wet epoxy. Salt crystals were rinsed off after the epoxy set.

31

Figure 3.1: Passive solar heated plate design. Top plate is either black or white HDPE with a central area of textured Sea Goin’® Poxy Putty where organisms settled. A hole in the underside of the top plate allows for a recessed iButton to directly measure epoxy temperature. The bottom plate was bolted to the rocky bench.

Settlement surface temperature was estimated using iButton temperature loggers (Thermochron DS1921G model, resolution ± 0.5°C; Maxim/Dallas Semiconductor Corporation, Dallas, Texas, USA), placed in a central recessed hole located under the epoxy surface, such that the sensor side of the logger was in direct contact with the epoxy. Loggers recorded the temperature once every 60 minutes. The plate design used two plates, the top one where organisms settled, screwed on to a second one, which was bolted into the intertidal rocky bench. This allowed the iButton to be sandwiched between plates (maintaining direct contact with the epoxy surface) and avoided protruding hardware on the settlement surface, which disproportionately encourages settlement (personal observation). The temperature of the surrounding rocky bench was monitored using iButtons embedded in Z-Spar (A-788 Splash Zone Compound) in each zone (n = 3), interspersed among plates. All plates were installed on a gently sloping (0-20°) southeast facing bench, although the high zone plates were on average slightly more horizontal than the rest due to topographic constraints at the site (ANOVA testing plate angle for treatment and intertidal zone: zone F2,11=5.191 p=0.026, temperature treatment

F1,11=0.129 p=0.726, temp × zone F2,11=2.585 p=0.120). Within each of three zones, seven plates of each color were affixed to the rocky bench in an irregularly alternating sequence, spaced 15-30 cm apart. Plates were installed between April 12 and 30, 2009 and monitored at two-week intervals during summer months (May-September) and approximately every three months during winter (September, November, March). At each

32 census date, densities of invertebrates and percent cover of all algal species on the epoxy surfaces were estimated. Understory algal cover was estimated by moving the canopy layers aside, thus algal estimates sometimes exceeded 100% cover. Barnacles were subsampled (31% of epoxy area counted) on plates. A 3 mm area at the edge of the epoxy was excluded from counts to avoid edge effects, as some organisms were observed to prefer the topographic heterogeneity provided by the raised epoxy edge. The black and white puckboard plate borders were also scrubbed clean of any diatoms and macroalgae at every sampling date to maintain temperature treatments.

3.3.3 Statistical analyses

Average daily maximum temperature was calculated for each plate and used in analyses. Due to logger failure, few individual plates had complete temperature records for the duration of the experiment, which could bias thermal comparisons among plates due to differences in the time periods over which temperatures were recorded. Thus, I calculated the residuals for each plate from the grand mean on each sampling date. The average residual across dates for each plate was then calculated, and used in an analysis of variance (ANOVA) with plate as the level of replication to determine if temperature varied among plate colors or shore levels. Alpha diversity and the abundances of barnacles, macroalgae and grazing gastropods in response to tidal elevation and the temperature manipulation were analyzed as a repeated measures design using linear mixed effect models (SAS Institute Inc 2009). Linear mixed model procedures offer two advantages over traditional ANOVA for the analysis of repeated measures designs. First, the covariance structure of the data among repeated observations can be modeled and accounted for directly. Second, mixed model procedures can accommodate missing observations in repeated measures designs, which occurred beginning with sampling in the fall 2009 and throughout spring 2010. The overall design consisted of a fixed effects model of temperature crossed with tidal zone and each of those factors crossed with date, the repeated measure. I employed Burnham and Anderson’s (2002) approach using Akaike Information Criteria (AIC) to infer the best supported model(s) of the various main effects and interactions. For most responses, 18 different models were tested, that each tested the importance of a particular source of variation to the outcome of a response variable. The analysis for total algae encompassed 36 different models, because each of the 18 models were tested with and without

33 the inclusion of grazer density as a covariate. Corrected Akaike Information Criteria values

(AICc) from analysis of each model were used to rank them from smallest to largest and compute their differences. Akaike weights (ωi) were calculated from the differences between AICcs (Δi) for each model and the model with the minimum AICc. The relative likelihood of the best-fitting model to the next best, or any other model(s) was calculated from the evidence ratio of respective model weights, ωi/ωj. Model parameters were fitted using restricted maximum likelihood (REML) estimation. Examination of the variance-covariance matrix among repeated observations indicated that modeling repeated observations using an unstructured covariance matrix was appropriate. Using REML to model effects in PROC MIXED assumes the data are normally distributed and two procedures were employed to meet this assumption. First, the Box-Cox transformation procedure was used to identify the best transformation to improve normality (given in Table B.1). For some variables, it was necessary to add 1 to each variate to avoid division by zero. The data for some variables remained leptokurtotic after transformation because of many zeros in a data set. To normalize those data, random noise was added to the data set by adding a value to each observation drawn randomly from a normal distribution with a mean equal to 0.1% of the maximum value of the variable observed in the data and a correspondingly small standard deviation so that all values of variates remained greater than zero. This procedure has no effect on the relative order of variates among groups, it only introduces normally distributed errors within groups to meet assumptions of the test and so is a conservative approach (P. Petraitis, pers. comm.). To test my hypothesis that community structure would be affected by temperature treatment, I analyzed the assemblage structure of the community with multivariate statistics, using PRIMER 6 software (Clarke and Warwick 2001). I used community data from the August 15, 2009 sampling date because it was the latest date before loss of replicates became prevalent. Percent cover of algae and invertebrate density were analyzed separately because different types of data cannot be combined. I square root transformed invertebrate density prior to applying the Bray-Curtis similarity coefficient to construct the resemblance matrix, to down weight similarity between blank or sparsely settled plates. Total algae percent cover was first log(x+1) transformed to down weight contributions of quantitatively dominant species (Clarke and Warwick 2001), before constructing an MDS plot using Euclidean distance. Permutational multivariate analysis

34 of variance (PERMANOVA) was used to examine variation in community assemblage between temperature treatments and zones. Each term in the analysis was tested using 9999 random permutations of the appropriate units. When significant results were detected (p<0.05), a SIMPER analysis was conducted to determine the principal contributors to dissimilarity in community composition. Finally, to determine if the community on top of the plates was influenced by mobile consumers sheltering under the plates, I quantified grazers and predators under plates during two census dates (July and September, 2009). Abundance of consumers found under plates was analyzed with a 3-factor ANOVA, testing the effects of temperature treatment, zone, and sampling date. I also visualized the assemblage using MDS plots, on Euclidean distance calculated from log transformed data, and used PERMANOVA to examine variation between temperature treatment, zone, and sampling date. Each term in the analysis was tested using 9999 random permutations of the appropriate units. Data are expressed in means ± standard error where applicable.

3.4 Results

3.4.1 Temperature

Temperature differences between treatments only existed during daytime low tides in the summer and not in winter when low tides were at night (Chapter 5). Subsequent experiments using the same method at the same location confirmed the similarity of treatment temperatures over winter months. Thus, summer temperatures drove the differences on plates. The temporal variation in average daily maximum (ADM) temperature (averaged across replicate plates) tracked the fortnightly tidal cycle in the low and mid zones, where plates remained completely submerged or minimally exposed for consecutive days during neap tide series (Fig. 3.2). Plates in the high zone were rarely submerged for more than several hours per day (except during extreme spring tides), hence had consistently high plate temperatures. Temperature was low (8- 15°C) and similar across different colored plates during high tides (Fig. 3.3). During daytime low tides, plates and the rock warmed by more than 20°C and black and white plates diverged (Fig. 3.2, Fig. 3.3). The temperature of white plates very closely approximated that of the surrounding bedrock (Fig. 3.4). ADM temperatures were consistently higher on black plates and at higher

35 shore levels during the summer months, with no significant interaction between treatment and zone (ANOVA: temp F2,30=38.847, p<0.001; zone F2,30=435.169, p<0.001; temp × zone

F4,30=1.423, p=0.250). The average ADM temperature difference between black (hereafter, warm) and white (hereafter, cool) plates in high, mid, and low zones was 2.9, 2.0, 1.5°C respectively (Table 3.1). The difference between the highest recorded temperatures of warm and cool plates were 4.0, 6.0, 5.0°C for the high, mid, and low zones, respectively (Table 3.1). Within each zone, there were more days reaching hot daily maximum temperatures on warm plates compared to cool plates and the surrounding bedrock (Table 3.1). For example, in the high zone, warm plates were “hot” (≥ 33°C) for 51 days and “very hot” (≥ 37°C) for ten of those days, whereas cool plates were hot for 17 days and were never very hot.

Figure 3.2: Average daily maximum summer temperature for (a) high, (b) mid, and (c) low zones measured by iButtons sampling at 60 min intervals. Blue symbols represent cool (white) plates, red symbols represent warm (black) plates, and black symbols represent the temperature of the nearby rocky bench. Error bars are ± standard error throughout.

36

Figure 3.3: Association between plate temperature and environment over a 5-day low-tide series in August, 2009. The average temperature (°C) of cool (open blue circles) and warm (solid red circles) replicate plates and the nearby rocky bench (open black squares) in the high intertidal are displayed on the left y-axis. Epoxy surfaces of plates and rock temperatures were sampled every 60 minutes using iButton loggers. Air temperature data (dotted black line) are from Environment Canada for nearby Victoria, BC. The right y-axis displays the tidal regime (solid blue line). Error bars are excluded here for clarity, but are included in time series data for the entire summer (Fig. 3.2).

Figure 3.4: Mean residual daily maximum temperature from plates and bench for each zone. Letters that are different than each other indicate significant differences at α = 0.05. Error bars are ± standard error throughout.

37

Zone Trmt Mean Mean Mean Mean Mean 25th – 75th % “Hot” days, “Very hot” days, min (raw) (ADM) max variance range ADM ≥ 33° ADM ≥ 37° High Black 6.0 15.5 29.9 40.9 53.6 6.5 51 10 White 5.4 14.8 27.3 36.4 39.7 6.1 17 0 Rock 7.8 15.4 26.6 37.5 35.2 5.0 21 3

Mid Black 7.6 13.3 22.7 37.8 18.1 2.1 19 6 White 7.2 12.6 20.8 33.8 13.7 2.1 4 0 Rock 9.1 13.3 21.7 32.4 15.7 2.4 1 0

Low Black 9.4 13.4 19.5 38.7 11.6 3.0 13 5 White 9.5 12.5 18.0 34.1 7.8 1.4 3 0 Rock 10.0 12.1 17.5 28.3 4.7 1.3 3 0

Table 3.1: Summary statistics of plate temperature and the number of days when the Average Daily Maximum (ADM) temperature, averaged across plates, was over 33°C and 37°C. All statistics were calculated using the raw temperature data except where noted (when ADM was used). 33°C was chosen as a significant threshold because the heat shock response for B. glandula is maximal at this temperature (Berger and Emlet 2007). 37°C was chosen as the “very hot” temperature threshold for illustrative purposes; it is the lowest temperature at which high cool plates never surpassed and is an important physiological threshold for barnacle parasites (Harley and Lopez 2003).

3.4.2 Functional group responses

The majority of algal cover quantified in this experiment was from prostrate species that laid in contact with the plate surface, so it is likely that thalli were warmed on black plates (Table A.1). Differences among tidal zones in the temporal dynamics of algal abundance depended on temperature treatment (Fig. 3.5d). In the high intertidal, cool and warm treatments showed similar patterns of algal abundance, peaking at 100% cover in August, which thereafter declined throughout fall to a minimum the following winter and spring (Fig. 3.5a). In contrast, warm temperatures in the mid and low zones delayed the development of algal communities from late spring into summer and algae persisted at greater abundance thereafter in the low zone (Fig. 3.5b-c). This three-way interaction model was more than 50 times more likely than any other model in the set, indicating that other models were essentially unsupported by the data (Table B.1a). Models including grazer density as a covariate were not better supported than the corresponding models lacking the covariate (not shown). Within a month after plates were installed, there was a large bloom of green algae (primarily Ulothrix spp and Urospora spp) in the mid and low zones, which died off by August

38 (Fig. 3.6). The green algae bloom occurred later in the high zone, peaking in September. Analysis of green algal percentage cover alone also supported the model including the three-way interaction, being 32 times more likely than other models (Table B.1b). In late summer and winter, species of reds, browns, and diatoms began colonizing both treatments of plates in the low and high zones (Fig. 3.6), although the prevalence of zeros in the datasets precluded analysis of these groups. Grazer densities were reduced in warmed treatments in all zones (Fig. 3.7). Grazers were often four to ten times more abundant in cool treatments than warm treatments, except in the low zone where differences diminished over time (Fig. 3.7d). The best model describing grazer abundance was one based on temperature treatment alone, which was more than seven times more likely than other models (Table B.1c). Grazers did not appear on plates until July, and consisted entirely of limpets (mostly Lottia pelta in the low and mid zones and L. digitalis and L. paradigitalis in the high zone) until winter, when some littorine snails (Littorina plena and L. scutulata) appeared on plates (Fig. 3.8). The difference in barnacle (Balanus glandula and Chthamalus dalli) density between treatments increased over time (Fig. 3.9). The pattern of barnacle abundance was best described by the model including the three-way interaction among temperature, zone, and date (Table B.1d). This model was at least 107 times more likely than any other model, providing overwhelming evidence that other models were unsupported. In late summer (July-August) and winter (January and March), there were large differences between treatments estimated between zones. For example, on July 19, barnacles were twice as abundant in cool treatments in the low and high zones, and 13 times more abundant on cool than warm in the mid zone (Fig. 3.9d). The magnitude of those differences between zones varied between dates (Fig. 3.9d).

39

Figure 3.5 (left): Response of algae to temperature treatments at each sampling date, over one year. Mean total percent cover of algae on cool and warm plates in the (a) high, (b) mid, (c) and low zones. Means exceed 100% due to layering of algae. (d) The relative difference in total algae percent cover between temperature treatments (ln(warm/cool)) for each zone. Positive values indicate a positive effect of warming, negative values indicate a negative effect of warming. Error bars are standard error. Note the difference in x-axis spacing in top graphs compared to (d).

Figure 3.6 (right): Mean percent cover of algal functional groups on cool plates (open symbol, dashed line) and warm plates (solid symbol, solid line) in (a) high, (b) mid, and (c) low zones. (d) The relative difference in green algae percent cover between temperature treatments (ln(warm/cool)) for each zone. Positive values indicate a positive effect of warming, negative values indicate a negative effect of warming.

40

Figure 3.7 (left): Response of grazers to temperature treatments at each sampling date, over one year. Mean density of total grazers on cool and warm plates in the (a) high, (b) mid, (c) and low zones. (d) The relative difference in grazer density between temperature treatments (ln(warm/cool)) for each zone. Positive values indicate a positive effect of warming, negative values indicate a negative effect of warming. Note the difference in x-axis spacing in top graphs compared to (d).

Figure 3.8 (right): Mean density of grazer functional groups: limpets and littorine snails, on cool plates (open symbol, dashed line) and warm plates (solid symbol, solid line) in (a) high, (b) mid, and (c) low zone.

41

Figure 3.9: Effect of temperature treatment on barnacle abundance in (a) high, (b) mid, and (c) low zones at each sampling date over one year. (d) The relative difference in barnacle density between temperature treatments (ln(warm/cool)) for each zone. Positive values indicate a positive effect of warming, negative values indicate a negative effect of warming. Note the difference in x-axis spacing in top graphs compared to (d).

42 3.4.3 Community response

Temperature affected species richness differently in each zone and over the duration of the experiment. The model that best described the pattern is one that included the three-way interaction between temperature, zone, and date and was at least 43 times more likely than all other models (Table B.1e). Species richness was generally reduced in warm treatments (Fig. 3.10). In early summer and winter, the difference between temperature treatments was higher at higher shore levels (Fig. 3.10d). During mid-late summer, species richness was 1.5-2.5 times lower on warm plates in the mid and low zones, while the difference between treatments in the high zone was negligible. In the mid and high zones, limpets and littorine snails were rare or absent on warm plates (Fig 3.8), accounting for much of the difference between treatments by the end of the experiment. On plate settlement surfaces, MDS ordination indicated partitioning due to temperature in the structure of the invertebrate assemblage in August (Fig. 3.11). I detected significant differences in multivariate assemblage structure between temperature treatments and zone for invertebrates (Table 3.2). Barnacles and limpets were the largest contributors to the dissimilarity between temperature treatments, and all were less abundant on warm plates (Table 3.3). The same taxa were also responsible for differences between zones; relative abundances contributed to 33-38% dissimilarity between pairs of zones (data not shown). MDS ordination indicated partitioning and PERMANOVA revealed significant differences in algal assemblage structure between zones but not between temperature treatments (Fig. 3.11, Table 3.2). Green filamentous algae, diatoms, and Ulva spp contributed to differences between zones (80-100% dissimilarity between pairs of zones). In September, the low zone was characterized by high diatom and Ulva spp coverage, the mid zone was characterized by bare plates (0% cover), while the high zone was characterized by high green filament cover and moderate diatom and Ulva spp cover.

43

Figure 3.10: Effect of temperature treatment on species richness in (a) high, (b) mid, and (c) low zones at each sampling date over one year. (d) The relative difference in species richness between temperature treatments (ln(warm/cool)) for each zone. Positive values indicate a positive effect of warming, negative values indicate a negative effect of warming. Note the difference in x-axis spacing in top graphs compared to (d).

44

Figure 3.11: MDS ordination plots of community composition for the (a) abundance of invertebrates (square root transformed with Bray Curtis similarity), and (b) percent cover of algae (log x +1 transformed with Euclidean distance) on the August, 15 2009 sampling date (n=40 plates). Filled symbols represent warm plates, open symbols represent cool plates in the high (triangles), mid (circles), and low (squares) zones. In the algae plot, many symbols are overlapping (e.g. in the center of the plot, where there was 0 % cover algae on plates).

Source df SS Pseudo-F P (a) Invertebrate density temp (t) 1 11262 9.749 <0.001 zone (z) 2 10737 4.647 <0.001 t × z 2 1562 0.676 0.708 residual 34 39279 total 39 62890

(b) Algae percent cover temp 1 3.53 1.937 0.094 zone 2 43.27 11.883 <0.001 t × z 2 5.43 1.492 0.146 residual 34 61.90 total 39 114.78

Table 3.2: Permutational multivariate analysis of variance of the effects of temperature treatment and zone for (a) the density of invertebrates (square root transformed, using Bray Curtis similarity) and (b) percent cover of algae (log(x+1) transformed, using Euclidean distance) for the August 15, 2009 sampling date. Bold p-values indicate significance at α = 0.05.

45

Species Warm (%) Cool (%) Diss / SD Contr. (%) Cum. (%) Chthamalus dalli 1.83 3.52 1.32 38.51 38.51 Balanus glandula 1.89 3.74 1.21 34.07 72.58 Barnacle recruits 1.00 0.83 1.26 13.25 85.83 Limpets 0.72 1.47 0.86 11.73 97.55

Table 3.3: Percentage contributions of individual species to observed differences between warm and cool plate invertebrate assemblages as estimated by SIMPER analysis. Only the top four contributors to dissimilarities are shown. Average dissimilarity between treatments was 42.75%. Warm: mean density on black plates; Cool: mean density on white plates; Diss/SD: average dissimilarity divided by standard deviation; Cont.: contribution of each species to differences between treatments; Cum.: running total of the contribution to the observed dissimilarity.

The total abundance of consumers (grazers + predators) found under plates did not differ between temperature treatments, but was greater in lower zones and in July (ANOVA: temp

F1,71=0.987, p=0.324; zone F2,71=24.443, p<0.001; date F1,71=14.660, p<0.001; date × zone

F2,71=5.195, p=0.008; date × temp F1,71=1.254, p=0.266; zone × temp F2,71=0.432, p=0.651; 3- way F2,71=0.224, p=0.800). Abundance of consumers was higher in July than September in the low and high zones, but was approximately equal between dates in the mid zone. I conducted similar analyses with limpets + chitons only, grazers alone, and predators alone, and found the same pattern of significance among treatments (data not shown). The results for the assemblage of consumers were similar to the results for abundance of consumers; MDS ordination indicated no partitioning due to temperature treatments (PERMANOVA: temp Pseudo-F1,71=0.505, p=0.899; Table 3.4), but did indicate partitioning in the structure due to zone and sampling date (Fig. 3.11).

46

Figure 3.12: Non-metric multi-dimensional scaling ordination plots for the log(x+1) Euclidean distance for community structure beneath plates (n=40 plates). Red symbols represent warm plates, blue symbols represent cool plates in the high (triangles), mid (circles), and low (squares) zones. Solid symbols are for the July survey and open symbols for the September survey.

Source df Pseudo - F P(perm) temp 1 0.505 0.899 zone 2 27.577 < 0.001 date 1 10.120 < 0.001 temp × zone 2 0.545 0.950 temp × date 1 0.419 0.943 zone × date 2 4.605 < 0.001 3-way 2 0.592 0.926 residual 71

total 82

Table 3.4: Permutational multivariate analysis of variance of the effects of three crossed factors (temperature treatment, zone, sampling date for log(x+1) abundance of organisms beneath plates. Bold p-values indicate significance at α = 0.05.

47 3.5 Discussion

Temperature is a fundamental determinant of survival and performance for all species. Because different species have different thermal tolerances and different morphological and behavioral coping strategies, species-specific responses to warming are inevitable. Interspecific differences provide one possible starting point for predicting temperature-driven changes in assemblage structure (Helmuth et al. 2005). In this intertidal system I used existing information regarding variation in thermal limits among taxa to make several predictions. For example, though I expected all taxa to be negatively affected by warming, I predicted that barnacles would fare well relative to limpets and algae. The upper thermal limit (in three hour aerial exposures) is over 44°C for Pacific high intertidal barnacles (Liao and Harley, unpublished data) and 39-44°C for Pacific limpets (Wolcott 1973). The upper thermal limits for species of prostrate intertidal macroalgae, measured in water, range from 20-33°C (Table A.1). Therefore, I hypothesized that barnacles would fare slightly better than grazers, and that the latter would avoid the hot microclimate created by warm treatments. Given that little is known about the thermal tolerances of northeast Pacific algal species in air, it was not possible to precisely predict how prostrate algae would fare compared to other functional groups. However, I expected that algae would be sensitive to warming and that if mobile grazers avoided hot microclimates, palatable macroalgae would be released from grazing. I found that warming negatively affected the density of barnacles in all zones throughout the year, thus it is likely that temperatures in warm treatments exceeded physiological tolerances of barnacles. Though much is known about the effect of temperature on barnacles in general (Foster 1969), less is known about northeast Pacific intertidal species. Balanus glandula heat shock protein expression begins at 28°C in air and peaks at 33°C (Berger and Emlet 2007). Chthamalus dalli cirri (feeding apparatus) beating reaches maximum levels at 28°C and ceases at 35°C in water (Southward and Southward 1967). In Australia, barnacle settlement and growth were inhibited when the substratum was warmed 2.2°C (Lathlean and Minchinton 2012), confirming that Australian barnacles are sensitive to slight increases in substratum temperature. Subsequent work in this system will elucidate the mechanisms controlling Pacific barnacle demographic responses to warming (Chapter 4).

48 Contrary to my expectations, algal abundance and algal community structure were relatively unaffected by temperature treatments. Previous field experiments mimicking aerial thermal stress on macroalgae have found that the responses are context specific (Allison 2004) and likely depend on complex interactions. In the high zone, algal settlement was delayed until September, which may have been due to higher thermal and desiccation stress, nutrient limitation from long emersion times, or differences in algal reproductive phenologies. In lower zones, green filamentous algae (the dominant algal group recruiting to my experimental plates) responded to warming via a delay in phenology. Interestingly, after an initial delay in establishment on warmed plates, green algae population dynamics followed the same pattern (Fig. 3.6b-c). Studies from a variety of ecosystems have also found warming induced colonization or phenology shifts (Parmesan and Yohe 2003). However, most have found that these changes advance with warming, whereas here, I observed a delay in the warm treatment. In most cases, life history transitions are triggered by warming in the spring. For ephemeral intertidal algae, the bloom may be triggered by early successional amelioration of stress. For example, there may have been delayed establishment of an early successional biofilm (Morelissen and Harley 2007) needed for green algae settlement, as has been suggested to be required for coral larvae (Webster et al. 2010). Thus, the prediction of advance or delay in phenology depends on what warm temperatures enable (e.g. flowering) or hinder (e.g. early succession). The average maximum on black plates was approximately 4°C warmer than white plates, and black plates stayed hot for longer within a day. Even if thermal limits were not exceeded by these conditions, sub-lethal stress may have caused mobile grazers to simply avoid this microclimate during low tide. Limpets (Lottia spp.) generally have high lethal thermal limits; 34- 39°C in water and 39-44°C in air (Wolcott 1973). Although these temperatures encompassed the warmest temperatures measured on the epoxy of my black plates, the black borders were likely a few degrees warmer, thus may have been too hot for limpets to cross during sunny low tides. Limpet adults and recruits were observed on black plate borders during low-tide surveys, although results from subsequent experiments revealed that limpets were approximately half as numerous on black HDPE borders than white borders during summer (Chapter 5). Nevertheless, limpets do not actively forage in hot conditions (e.g. daytime low tides; Branch 1981) and they had access to the ocean- or air-cooled epoxy surface during high tide and at night. Thus, hotter

49 black borders may have restricted but not prevented limpet access to the epoxy surface. Regardless of the mechanism, warmer conditions reduced limpet abundance, which likely reduced grazing pressure to warmed intertidal communities. Warming effects on algae and limpets appeared to be primarily direct, rather than mediated indirectly by top-down or bottom-up forcing. This was evidenced by the gap between green (i.e., palatable) algae die-off in late summer and grazer appearance shortly thereafter and the late summer disappearance of algae on warm plates, which lacked grazers. Harley (2006) also observed a green algae die off in late summer in treatments where herbivores were removed. It is probable that green algae were an early successional functional group that allowed for grazers to either recruit to plates or facilitated a microbial food source that is preferable to grazers. However, there were fewer grazers in warm treatments; this difference was particularly evident in the mid and low zones, where limpets were numerically abundant, compared to the high zone. Thus, despite non-discriminatory ‘preparation’ by green algae, mobile grazers were less abundant on warm plates. Therefore, the difference in limpets on plates is likely due to abiotic conditions rather than availability of food. Subsequent studies controlling for the effect of grazing will shed light on the effect of warming on these interactions (Chapter 5). In summary, I found that functional groups responded differently to experimental warming; barnacles and limpets were reduced, and green algal phenology was shifted. Although laboratory-based physiological thermal tolerances can provide useful mechanistic insight into species responses to climate warming (Helmuth et al. 2005), the current shortage of data hinders our ability to apply this framework at the community level. Since species thermal responses are idiosyncratic, temperature performance data for a greater diversity of species (especially interacting pairs), physiological and behavioral responses, and habitat conditions (e.g. low tide and high tide), will provide climate change science with an essential database to build predictive models. The overall outcome of the taxon-specific changes detailed above was a reduction in richness and a change in invertebrate community structure with warming. This was due to the loss of mobile herbivores and reduction in barnacles in warm treatments, and by the failure of warm-adapted taxa to replace those losses. It is possible that my experiment overestimated species loss and underestimated species replacement due to the spatial and temporal scale of the manipulations. The experiment did not affect enough individuals over a long enough period of

50 time to promote local adaptation based on existing genetic variation; such adaptation could ameliorate abundance declines and reduce species losses. Furthermore, the manipulations did not affect water temperature, and non-resident warm-adapted taxa that require warmer water, e.g., for larval development, would not be expected to appear on the plates. Nonetheless, there is reason to believe, at least at a coarse level, that the decline in species richness I measured will reflect actual future change in this system. The Salish Sea (inland waters of Washington and British Columbia) is a thermal hot spot within the northeast Pacific, and summer average daily maximum temperatures have warmed by 3.4°C in the past 50 years (Harley 2011). The nearest shores likely to harbor species adapted to warmer conditions can found more than 1500 km distant, in southern California. There is very little variation in the species pool along the intervening coastline (Blanchette et al. 2008), so the availability of warm-adapted species to this system may be very limited. The Salish Sea may thus represent a system that is on the declining (stressed) portion of its thermal performance curve, with little prospect for the arrival of warm- adapted species in the foreseeable future. Contemporary responses of whole ecosystems to global warming have been context dependent (reviewed by Doney et al. 2012) and depend on a myriad of factors: specifically, the interplay between individual thermal tolerances and adaptations, context-dependent species interactions, additional abiotic stressors, rates of immigration, and the scale of observation. The paleontological literature reports massive declines in biodiversity followed by a global reshuffling of species (reviewed by Barnosky et al. 2012). Decadal-scale observations have demonstrated increased diversity and changes in community composition with warming (e.g. Stachowicz et al. 2002, Schiel et al. 2004, Perry et al. 2005), while shorter term experimental studies have demonstrated declines in diversity (e.g. this study; Gedan and Bertness 2009, Burgmer and Hillebrand 2010). Improving our understanding of the ultimate ecological responses to anthropogenic warming will involve linking mechanistic drivers to long-term, macro-scale patterns; conducting more field-based warming experiments will be one key step towards this goal.

51 Chapter 4 Demographic Responses of Coexisting Species to In Situ Warming

4.1 Synopsis

Climate warming may drive organismal body temperatures beyond physiological thresholds, ultimately leading to detrimental effects on populations and communities. Integrating warming effects across ontogeny in a natural setting can elucidate mechanisms behind warming responses that cannot be perceived by studying a single life stage. Much of what we currently know about species responses to warming has come from correlations with weather patterns or laboratory experiments, which can lack mechanism and realism, respectively. I incorporated both properties into warming experiments by manipulating substratum temperature in situ, using passively warmed black and white settlement plates. I monitored vital rates of coexisting barnacles over one year in the mid and high zones of the rocky intertidal on Salt Spring Island (British Columbia, Canada); a ‘hot spot’ for intertidal thermal stress. Warming by ~2°C negatively affected both species of barnacles, however the population of the competitive dominant, Balanus glandula was more severely affected than Chthamalus dalli, leading to a temperature-induced change in space occupancy. Survivorship of B. glandula was reduced by 87-100% leading to a 82-92% smaller population size and 89-94% lower space occupancy, depending on shore level. C. dalli survivorship was also reduced by warming (~50%) leading to a 62-80% smaller population size and 67-72% reduced space occupancy. Further, growth rates of C. dalli were lower in warm treatments resulting in 25% smaller adult body sizes. Though the effects on C. dalli population size were less severe, sublethal thermal stress likely caused delays in reproductive maturity, further reducing the fitness of the population. A mechanistic understanding of species-specific responses to warming is essential for more robust predictive models. Experiments like this one, which manipulate warming in the field on multiple species across ontogeny, will further enhance our understanding of the effects of climate change.

52 4.2 Introduction

Global warming has already resulted in pervasive alterations to natural systems (Walther 2002, Root et al. 2003, Buckley et al. 2011). Fundamental characteristics of populations, including abundance, spatial extent, the timing of demographic transitions, and body size frequency distributions are all changing in response to global warming (Pounds et al. 1999, Sagarin et al. 1999, Parmesan and Yohe 2003, Daufresne et al. 2009). As carbon emissions continue in the coming decades, global mean temperature will continue to increase, as will the frequency and magnitude of high-temperature events (IPCC 2013), intensifying changes to natural systems. Understanding how and why different species respond to temperature in unique ways will improve our ability to predict species’ population responses to climate change. To do this, predictions have often relied upon bioclimatic envelope models or correlations between populations and coarsely measured environmental factors (Poloczanska et al. 2008, Barbraud et al. 2010, Dalgleish et al. 2011). Recent work has highlighted that predictions can be improved by adding mechanistic information such as thermal sensitivity of vital rates and ontogenetic stages (Helmuth et al. 2005, Buckley et al. 2011). One approach is to incorporate correlations between species-specific physiological rates and long-term weather patterns (Walther 2002, Root et al. 2003, Peery et al. 2011). This represents an improvement insomuch as responses more specific than abundance are considered, but still relies on correlations to broad-scale environmental variables and often only considers a single life stage. Populating models with temperature- dependent processes through ontogeny, and derived from field-based manipulations, will lead to more accurate predictions, although this is not often possible due to the paucity of relevant data (IPCC 2007, Buckley et al. 2011, Huey et al. 2012). Barnacles are excellent model organisms for determining how temperature-specific vital rates can affect performance in a changing climate. Intertidal barnacles are marine organisms but spend much of their lives in terrestrial environments. Intertidal adults brood their embryos until their release as swimming naupliar larvae. When returning larvae, now at the cyprid stage, settle on intertidal substratum, they quickly secrete a shell and metamorphose into juveniles within a few days. This contact with the substratum is important for two reasons. First, it is permanent; the barnacle will remain in this location for the remainder of its life. Second, the relatively large contact area with the rock, combined with the barnacle’s small size, constrains barnacle body

53 temperature to within a few degrees of substratum temperature. The lack of mobility and dearth of effective behavioral thermoregulatory options coupled with the high variability in intertidal rock temperatures (e.g. Harley and Helmuth 2003) mean that barnacle body temperatures can fluctuate considerably – fluctuations of more than 20°C over a single low tide are common (CDGH unpublished data). For this study, I focused on Balanus glandula and Chthamalus dalli, which are co- occurring barnacles in the northeast Pacific that play an important ecological role in intertidal habitats. Adult barnacles can facilitate other intertidal species by increasing topographic heterogeneity, reducing desiccation and consumption of spores and larvae (Barnes 2000, Harley 2006). However, maturing barnacles can also negatively affect neighboring sessile species (Connell 1961b, Kordas and Dudgeon 2011). As a barnacle grows in diameter, the plates of its shell (i.e., ‘test’) dislodge or weaken the tests of neighboring barnacles or the holdfasts of macroalgae, or they simply overgrow organisms such as encrusting algae (Dayton 1971). Balanus glandula is faster growing and reaches a larger adult size than C. dalli. Thus, in the absence of predators, B. glandula outcompetes C. dalli for space (Dayton 1971, Menge 2000, Harley and O’Riley 2011). Presumably, because it is larger (or energetically more profitable) than C. dalli, B. glandula is the preferred barnacle prey by predatory whelks and sea stars (Paine 1966, Dayton 1971). This predator-induced release from competition is at least partially responsible for allowing C. dalli to persist in a space-limited habitat (Dayton 1971, Harley 2011). Barnacles are sensitive to warming, however most of what is known about barnacle thermal sensitivity comes from correlations with sea surface temperature or laboratory experiments with warmed water. Comparatively less is known about how these species respond in situ to aerial thermal stress, and very little is known about how different life stages of these species respond to temperature. Chthamalus dalli development and feeding (Southward and Southward 1967, Miller and Carefoot 1989), and B. glandula development, cyprid size, and fecundity (Barnes and Barnes 1956, Hines 1978, Pfeiffer-Hoyt and McManus 2005, Emlet 2006) are sensitive to changes in water temperature. A less extensive literature on aerial thermal stress suggests that it is important for barnacles, and sensitivity varies between species (Foster 1971). Changes in thermal sensitivity through ontogeny have not been explored for Pacific species, however, young barnacles (cyprids) are less tolerant of aerial warming than adult Semibalanus

54 balanoides (an Atlantic species; Crisp and Ritz 1967). For B. glandula, body temperatures of 33°C at low tide – a temperature frequently reached in the high intertidal during the summer – produce maximal levels of heat shock proteins (Berger and Emlet 2007), indicating that this is an important thermal threshold. When acclimated to higher temperatures, B. glandula failed to shift its heat-shock induction temperature, indicating that this species may be operating at maximal levels and unable to acclimate to higher temperatures (Berger and Emlet 2007). Further, following a half century of warming air temperatures in the northeast Pacific, the upper limits of B. glandula and C. dalli have moved down-shore (Harley 2011), implying that aerial thermal stress may have profound consequences for their distribution and abundance. Climate change will have complex and potentially negative consequences for these barnacles, but warming may affect these co-occurring species differently. Balanus glandula inhabits the high intertidal from the Alaskan Aleutian Islands to northern Baja Mexico (Barnes and Barnes 1956). It reproduces in the winter, with a major release of brooded larvae in spring, though there is often continuous brooding throughout the year or a second large release in the autumn (Barnes and Barnes 1956, Dayton 1971, Broitman et al. 2008). Prolonged reproduction is advantageous because some cyprids may settle after summer when conditions can become thermally stressful. The upper limit of C. dalli extends approximately 5 cm higher than B. glandula, where they co-occur in the intertidal zone in British Columbia (CDGH unpublished data) and the geographic range of C. dalli is slightly more northerly; from northern Alaska to southern California (Southward and Southward 1967). Chthamalus dalli also reproduces in the winter, but unlike B. glandula, only releases one brood starting in early summer (Dayton 1971). The upper lethal temperature of C. dalli is slightly higher than B. glandula (Liao & Harley, unpublished data), consistent with longer emersion times experienced by the former. I hypothesized that warming would affect the populations of these competing barnacle species differently due to differential thermal sensitivity in vital rates. I measured the responses of barnacles to conductive thermal stress in the field using passively warmed settlement plates in the mid and high intertidal. Vital rates (growth and survivorship) were monitored over one year to determine species-specific sensitivities across the ontogeny of barnacle benthic stages. In general, I predicted that younger barnacles would be more sensitive to thermal stress than older barnacles, as has been found for Atlantic species (Crisp and Ritz 1967, Foster 1969). Further, cyprids that settled later in the summer, when it is warmer, would fare worse than earlier settling

55 cyprids. I also predicted that growth and survivorship would be more sensitive to thermal stress in B. glandula than C. dalli, which could weaken the competitive dominance of B. glandula by reducing population density and / or space occupancy.

4.3 Methods

4.3.1 Study system

This experiment was conducted on the southeastern shore of Salt Spring Island, one of the southern Gulf Islands in British Columbia, Canada (48.753° N, -123.388° W). Tides around the Gulf Islands are mixed semidiurnal and had a maximum tidal range of 3.9 m during 2009-2010. Lower low tides in summer months (March-August) coincide with the hottest portion of the day (10am-3pm). Like nearby San Juan Island, which is climatically similar, Salt Spring Island can be considered a ‘hot spot’ (Helmuth et al. 2002) for stressful intertidal temperatures relative to the outer west coast. These conditions make Salt Spring Island an ideal location for the study of thermal stress on intertidal invertebrates.

4.3.2 Experimental warming treatment

I used black- and white-bordered settlement plates to assess the effect of warming on barnacles, at each of two tidal heights (1.6 m and 2.6 m above the lowest astronomical tide estimated by the Canadian Hydrographic Service chart datum) in the rocky intertidal in 2009-2010. Seven plates of each color were installed in each zone. The mid zone (1.6 m) was dominated by barnacles, limpets, mussels, Mastocarpus papillatus, Fucus gardneri, and Ulva spp. The high zone (2.6 m) consisted almost entirely of barnacles, littorine snails, and limpets. The experimental design is the same used for results of Chapter 3, though only results from the high and mid zone are presented here. Settlement plates were made of black or white High-Density Polyethylene (HDPE). Each HDPE plate was 15.25 × 15.25 cm with a centered 6.9 × 6.9 cm area of white epoxy (see Chapter 3 and Fig. 3.1; for details). All plates were installed on a gently sloping (0-15°) southeast facing bench and were adhered in a haphazard black-white sequence, spaced 15-30 cm apart.

56 Because stressfully cold temperatures for barnacles were not reached during this experiment (C. dalli are active at water temperatures below zero; Southward and Southward 1967), I focused my analyses on the physiologically stressful warm temperatures. Daily maximum temperatures were calculated for each plate. Temperature differences between treatments only existed during day time low tides in the summer and not in winter when low tides were at night (Chapter 5). Subsequent experiments using the same method confirmed the similarity of treatment temperatures over winter months. Due to logger failure, few individual plates had complete temperature records for the duration of the experiment. This could potentially bias thermal comparisons among plates due to differences in the time periods over which temperatures were recorded. To circumvent this bias, I calculated the residuals for each plate from the grand mean on each sampling date. The average residual across dates for each plate was then calculated, and these values were used in a 2-way analysis of variance to determine if temperature varied among plate colors or shore levels. The number of days when plate and rock temperatures reached or exceeded important thermal thresholds for barnacles were calculated based on the average temperature of replicate plates within each treatment on each day. Threshold temperatures of 33°C and 37°C were used because the former is the temperature at which B. glandula produces maximal levels of heat shock proteins (Berger and Emlet 2007), and 37°C is the thermal tolerance limit for barnacle parasites (Harley and Lopez 2003). The total number of days exceeding these thresholds were compared among treatments and zones using Pearson χ2 goodness of fit test. Finally, temperature variability was estimated by calculating the variance of residual temperatures for each plate (4-5 replicate records per treatment combination). A 2-way ANOVA was used to determine whether the temperature treatment of plates or zone affected variance in temperature. Plates were installed between April 12 and 30, 2009 and monitored at two week intervals during summer months (May-September) and every three months during winter (September- March). At each census date, live and dead barnacle densities were randomly subsampled (31% of epoxy area counted). Recently dead barnacles that still had shells (empty ‘tests’) intact on the substratum were counted to estimate mortality. A 3 mm area at the edge of the epoxy was excluded from counts to avoid edge effects, as barnacles were observed to prefer the topographic heterogeneity provided by the raised epoxy edge.

57 4.3.3 Statistical analyses

To determine how the temperature treatment and zone affected the abundance of barnacles, I used Repeated Measures Analysis of Variance (RM-ANOVA) to take time correlations across sampling dates into consideration. Live and recently dead barnacles were treated as separate response variables. Loss of replicates became prevalent starting in September, so only data through August 15, 2009 were used (n=24 plates). To check the sensitivity of excluding winter data in RM-ANOVA, I re-ran the analysis with all sampling dates (n=15) and the results were qualitatively similar. In the winter, I expected no temperature difference between plate colors (because low tides were at night), however it was possible that biological differences existed on the plates in winter as a result of carry-over effects from the summer. Therefore, I conducted a 2- factor ANOVA on the last sampling date (March 24, 2010) to test the main effects of temperature treatment and zone at the end of winter (n=15 plates). Live and recently dead barnacle density estimates were log transformed to achieve homoscedasticity. If the sphericity of the variance-covariance matrix was violated for RM-ANOVA, the degrees of freedom were altered according to the Mauchly test using Huynh-Feldt Epsilon. The analysis for live barnacles was repeated with each species, separately, to determine the relative influences of zone and temperature. Photographs were taken of all plates at all sampling dates and were later used to determine percent cover, survivorship, and growth rates of barnacles. To directly compare how B. glandula and C. dalli responded to temperature, and to detect a signal of competition, I quantified the percent cover of barnacles. Menge (2000) noted that percent cover is a better indicator of space limitation by barnacles than density, and can reflect the relative dominance of each species. In the absence of settlement, increasing percent cover would indicate increasing barnacle size, thus competition could be inferred by the simultaneous increase in percent cover of one species and decrease in percent cover of the other, and confirmed by observing test boundaries in photographs. I estimated percent cover of each species on each plate at each sampling date by overlaying the settlement area with a grid of 100 intersecting points. The species found under each point was noted and percent cover estimated. When barnacles were older, a single barnacle often occupied several points. Data were log transformed to meet assumptions of parametric tests. I tested whether percent cover was affected by temperature

58 treatment, zone, or species using RM-ANOVA on summer data and a 3-way ANOVA on the last date. I also followed the survivorship of 25 barnacles on each plate on a subset of dates in both zones. Life stage (cyprid or newly metamorphosed) and species (when identifiable) were recorded. June 5, August 4, and September 15, 2009 were chosen as the intervals because they encompassed the warmest part of the summer (intermediate dates during this period were excluded because a transient pulse of green filamentous algae prevented identification of barnacles in the mid intertidal). The total survivorship relative to the initial time point was calculated for each plate and for each date. Cyprids were difficult to identify to species, so species were pooled to compare survivorship of cyprids and metamorphosed barnacles in treatments. Cyprid survivorship was extremely low (and was less than 10% of survivorship of newly metamorphosed barnacles), particularly by September, so the analysis of the June-August period only is presented. Data were log transformed to meet assumptions of normality. A 2- factor ANOVA was used to compare the main effects of zone and temperature treatment. Because I suspected that survivorship varied by species, I also analyzed the survivorship for metamorphosed barnacles, for which species identification was reliable. Data were converted to proportions and transformed using arcsine square root prior to analyses. I tested whether survivorship varied with zone or temperature treatment in a RM-ANOVA, using the June-August and August-September survivorship periods. The same analysis was repeated with each species, separately, to determine the relative influences of zone and temperature. To determine growth rates, the basal area of barnacles was measured at each sampling date, using Image J 1.44v (National Institutes of Health, USA). I followed the size trajectories of nine barnacles of each species (per plate) that survived until the final sampling date. To compare growth rates across zones and species, the first two months of barnacle growth after settlement on the plates was isolated (regardless of when barnacles settled), and daily growth rates were calculated. These data were separated by species and into summer and winter settlers to account for seasonal effects on growth rates. On many plates, I was unable to locate nine barnacles because barnacle mortality was high (e.g. B. glandula high zone, black plates). Replication within treatment was sometimes reduced because plates were lost (especially in winter), or dense algae obscured barnacles in photographs in some treatments. Barnacle growth rates were averaged to generate a single value for each plate to avoid pseudoreplication. Plate-based

59 averages were log transformed to meet assumptions of normality and homoscedasticity. An analysis of covariance, using initial size as a covariate, was used to test the effects of temperature treatment and zone on growth rate. The effect of initial size was not significant (p > 0.25) in all instances (likely due to the similarity in initial barnacle sizes), so the term was dropped for ease of interpretation; the results of both analyses were qualitatively similar. Within a season, growth rates may vary depending on the timing of biotic or abiotic conditions (e.g. food availability or temperature) or barnacle age. To tease apart how warming influenced these factors, I conducted several complementary analyses on the C. dalli high zone data set (the only data set complete enough for this level of scrutiny). I isolated three cohorts that had at least three C. dalli individuals per plate from the barnacle size database (described in the previous paragraph): those that settled just prior to May 22, June 24, and July 9, 2009, and calculated their daily growth rates (mm2 day-1) through March 2010. The average growth rate per plate was used to test the effects of temperature treatment and settlement date with RM- ANOVA. In the first analysis, I standardized for age by only using data from the first eight weeks of growth for each cohort of settlers. In a second analysis, I looked at how growth rates differed during the same time of year (July 19 – Sept. 15), for the three cohorts of settlers (i.e. for different ages). Analogous analyses were performed on C. dalli size data. Finally, I conducted an ANOVA on the last date to determine if the body size of adults was affected by summer warming. Data were log transformed to meet assumptions of normality and homoscedasticity. Statistics were performed in JMP Pro 9.0.3 (SAS Institute).

4.4 Results

4.4.1 Temperature

The daily maximum temperature, averaged across plates (average daily maximum, ADM) varied through the fortnightly tidal cycle in the mid zone and, to a lesser degree, in the high zone. Mid zone plates (1.6 m height) reached higher temperatures (ADM 25-38°C) during spring tides that coincided with the hottest portion of the day (10:00-15:00), and lower temperatures (ADM 10- 20°C) during neap tides when plates were emersed for fewer than four hours per day (Fig. 3.2). High zone plates (at 2.6 m height) stayed relatively warmer (ADM 25-42°C) through most of the summer because they were only submerged during the higher of the two high tides, which were

60 at night during spring and neap tides. Because temperature differences were created by solar radiation, all plates were approximately the same temperature at high tide and at night (Fig. 3.3), when they remained cool (mean minimum: 5-9°C; Table 3.1). Ocean temperatures were typically 8-12°C, but air temperatures as low as 4.5°C were reached during low tide at night in the spring. Once exposed to the sun at low tide, plates warmed by more than 20°C and the temperature on black and white plates diverged. ADM temperatures were consistently higher on black plates and at higher shore levels during summer months (temperature treatment: F2,21=34.674, p<0.001; zone: F1,21=327.648, p<0.001; temperature × zone: F2,21=1.694, p=0.208). The difference in ADM temperatures between black (hereafter, warm) and white (hereafter, cool) plates in the high and mid zones was 2.6 and 1.9°C respectively, and the temperature of white plates very closely approximated that of the surrounding bedrock (Table 3.1). There were more days reaching hot daily maximum temperatures on warm plates compared to cool plates and the surrounding bedrock, and more hot days higher in the intertidal (χ2 goodness of fit: temperature treatment: χ2 = 41.646, p < 0.001, zone: χ2 = 37.389, p < 0.001, Fig. 4.1, Table 3.1). For example, in the high zone, warm plates were “hot” (ADM ≥ 33°C) for 51 days and “very hot” (ADM ≥ 37°C) for ten of those days, whereas cool plates were hot for 17 days and were never very hot. In the mid zone, plates and the rock exceeded these thresholds half as often (or less) relative to comparable treatments in the high zone. Temperature variability differed across temperature treatments, but post hoc tests revealed that warm and cool plates were not significantly different (Table 4.1a). However, when black and white plates were compared directly without including a third ‘rock’ treatment, in order to match the biological comparisons described below, the thermal variance was significantly higher on black plates than on white plates (Table 4.1b).

61

Figure 4.1: Frequency histograms of the average daily maximum (ADM) temperature on plates and nearby rock for the high (top set) and mid (bottom set) zones. Temperatures are averaged from 1-3 loggers, sampling at 60 minute intervals, for each treatment combination. Loggers were deployed from April-September 2009. The vertical dotted line at 33°C represents an important thermal stress threshold for B. glandula (Berger & Emlet 2007). Sample sizes indicate the number of days with temperature readings.

62 Source df SS F P (a) black plates vs. white plates vs. rock temperature (t) 2 83.278 4.705 0.020

zone (z) 1 29.546 3.338 0.082 t x z 2 43.728 2.470 0.109 error 21 185.856 total 26 333.334

(b) black & white plates only temperature 1 53.913 6.325 0.024 zone 1 9.351 1.097 0.312 t x z 1 36.610 4.295 0.056 error 15 127.848 total 18 218.830

Table 4.1: ANOVA for variance in temperatures of (a) plates and rock and (b) plates only for Imid and high zones. The variance was calculated from residual temperature data over the summer months, for each plate or iButton in rock, and used in the analysis. N=27 temperature records (4-5 replicate records per treatment combination).

4.4.2 Barnacle abundance

The abundance and space occupancy of B. glandula and C. dalli populations varied over time and vertical space in the intertidal (on plates averaged across temperature treatments; Fig. 4.2- 4.3). Early patterns of barnacle recruitment (defined as the number of individuals that settled and survived for long enough to be counted at the end of the 23 day settlement period), assessed on May 22, 2009, revealed that recruitment was higher in the mid zone (zone: p<0.001, Table 4.4b) and almost twice as high for B. glandula compared to C. dalli (spp: p=0.013, Fig. 4.2, second time point). Over the course of the summer, the abundance and space occupancy of each species generally increased through time, but the magnitude of this increase, and the degree to which it was influenced by zone, differed between the two species (Density: spp × zone × time: p<0.001, Table 4.2, Fig. 4.2; Percent cover: spp × zone × time: p<0.001, Table 4.3, Fig. 4.3). In the mid zone, B. glandula was numerically dominant (Fig. 4.2b,d) and occupied more space than C. dalli (Fig. 4.3b). As the summer progressed, B. glandula cover increased, while cover of C. dalli remained low (Fig. 4.3b). In the high zone, C. dalli were numerically dominant (Fig. 4.3a,c), especially in late summer, but the two species occupied the same amount of space (Fig. 4.2a). By March, population sizes were equivalent between species and intertidal zone (for both density and percent cover: spp: p>0.200, zone: p>0.200, Table 4.2c, 4.3b). Both barnacle species were less numerically abundant on warm plates compared to cool plates. Barnacle recruitment on warm plates was half that on cool plates, although the trend was

63 not significant (temperature: p=0.086, Table 4.2b, second time point in Fig. 4.2) because densities were variable in the mid zone and low but similar in the high zone. Warming reduced barnacle density by an average of 84% overall (temperature: p<0.001, Table 4.2, Fig. 4.2), and the reduction was similar for both species (temperature x spp, temperature x spp x time: p>0.100) and in both zones (temperature × zone, temperature × zone × time: p>0.100). The difference between treatments varied over the summer; when barnacle density declined (on June 24 and August 4), it did so more sharply in warm treatments (temperature × time: p<0.001, Fig. 4.2). After one year, barnacle density was 90% lower in warm treatments (temperature: p=0.005) but was otherwise similar across species and zones (Table 4.2c, Fig. 4.4). Percent cover of barnacles was 80% lower in warm treatments overall (temperature: p<0.001, Fig. 4.3; Table 4.3a). Cover of barnacles remained low on warm plates throughout the year (under 10%), while cover increased over time on cool plates (temperature × time: p<0.001). Temperature affected the cover of each species differently (temperature × spp: p<0.001, see Table C.1 for relative temperature effects over the entire experiment). Although barnacle cover increased more rapidly on cool than warm plates for both barnacle species over summer, the divergence was more pronounced for B. glandula than C. dalli (temperature × spp × time: p=0.001), and the difference was most pronounced in the mid zone (temperature × zone × spp: p=0.025). On cool plates, B. glandula occupied more space than C. dalli, and on warm plates the two species occupied the same (minimal) amount of space (Fig. 4.3). The large apparent late- summer drop in B. glandula percent cover on cool plates in the mid zone was due to the loss of one particularly densely settled plate in September (Fig. 4.3), however this did not affect statistical results because only sampling dates through late August were used in repeated measures ANOVA, due to greater loss of replicates starting in September. By March, the percent cover of barnacles was 88% lower in warm treatments (temperature: p<0.001), regardless of species or zone (Table 4.3b).

64

Figure 4.2: Mean density and standard error (per 43.56 cm2 plate) of (a,b) B. glandula and (c,d) C. dalli in each of high and mid zones from April 2009 to March 2010. Averages and error limits were back-transformed for all graphical presentations.

65

Figure 4.3: Mean percent cover of barnacles per plate in the (a) high zone and (b) mid zone over one year. The apparent drop in B. glandula percent cover in the mid zone in September stems from the loss of one densely settled cool plate.

66

Source df SS F P (a) RM-ANOVA (through Aug 2009) n=48 Between-subjects temperature (t) 1,40 29.427 <0.001 zone (z) 1,40 1.496 0.228 species (spp) 1,40 0.352 0.556 t x z 1,40 0.928 0.341 t x spp 1,40 2.781 0.103 z x spp 1,40 7.440 0.009 t x z x spp 1,40 0.092 0.763 Within-subjects time 6,240 7.893 <0.001 temp x time 6,240 4.364 <0.001 zone x time 6,240 18.464 <0.001 spp x time 6,240 2.773 0.013 t x z x time 6,240 1.531 0.169 t x spp x time 6,240 0.710 0.642 z x spp x time 6,240 4.532 <0.001 t x z x spp x time 6,240 0.679 0.666 Mauchly criterion = 0.428, df = 20, P = 0.043

(b) ANOVA (May 22, 2009) n=52 temp 1 1.163 3.075 0.086 zone 1 8.098 21.409 <0.001 spp 1 2.532 6.695 0.013 t x z 1 0.408 1.077 0.305 t x spp 1 0.348 0.921 0.343 z x spp 1 0.011 0.028 0.868 t x z x spp 1 0.002 0.007 0.936 error 44 16.643 total 51 28.848

(c) ANOVA (March 2010) n=30 temp 1 3.446 9.57 0.005 zone 1 0.494 1.367 0.255 spp 1 0.536 1.483 0.236 t x z 1 0.217 0.600 0.447 t x spp 1 0.306 0.848 0.367 z x spp 1 0.400 1.109 0.304 t x z x spp 1 0.237 0.656 0.426 error 22 7.948 total 29 17.174

Table 4.2: Density of barnacles over one year. Repeated Measures ANOVA for (a) summer data and ANOVA on (b) first and (c) last sampling date. Degrees of freedom (df) listed as numerator, denominator for RM-ANOVA. Bolded values are significant at α = 0.05. Data were log transformed prior to analyses. n refers to the number of replicate plates. Degrees of freedom & P- value adjusted by Huynh-Feldt ε in RM-ANOVA.

67

Source df SS F P (a) RM-ANOVA (through Aug 2009) n=48 Between-subjects temperature (t) 1,40 48.734 <0.001 zone (z) 1,40 14.368 <0.001 species (spp) 1,40 6.713 0.013 t x z 1,40 4.138 0.049 t x spp 1,40 14.099 <0.001 z x spp 1,40 15.520 <0.001 t x z x spp 1,40 5.447 0.025 Within-subjects time 5.6,225.2 12.299 <0.001 temp x time 5.6,225.2 9.934 <0.001 zone x time 5.6,225.2 5.252 <0.001 spp x time 5.6,225.2 0.454 0.831 t x z x time 5.6,225.2 1.536 0.172 t x spp x time 5.6,225.2 3.857 0.001 z x spp x time 5.6,225.2 4.046 <0.001 t x z x spp x time 5.6,225.2 2.150 0.053 Mauchly criterion = 0.181, df = 20, P <0.001

(b) ANOVA (March 2010) n=30 temp 1 3.298 25.051 <0.001 zone 1 0.016 0.122 0.730 species 1 0.020 0.152 0.700 t x z 1 0.058 0.444 0.512 t x spp 1 0.004 0.029 0.865 z x spp 1 0.002 0.012 0.915 t x z x spp 1 0.098 0.743 0.398 error 22 2.897 total 29 8.616

Table 4.3: Percent cover of barnacles over one year. Repeated Measures ANOVA for (a) summer data and ANOVA on (b) last sampling date. Degrees of freedom (df) listed as numerator, denominator for RM-ANOVA. Bolded values are significant at α = 0.05. Data were log transformed prior to analyses. n refers to the number of replicate plates. Degrees of freedom & P-value adjusted by Huynh-Feldt ε in RM-ANOVA.

68

Figure 4.4: Representative photographs of warm (black) and cool (white) plates and the nearby rocky bench (cleared of organisms on the same day as plate deployment) one year after the start of the experiment. Dark brown-grey barnacles are Chthamalus dalli, white-to-tan barnacles are Balanus glandula. In the bottom photograph, the wire quadrat is the same size as the epoxy surface of the plates

69 Over the summer, B. glandula density differed between temperature treatments and zone, but these distinctions disappeared by the final sampling date. In contrast, cover remained lower on warm than cool plates throughout the experiment, and summer differences between zones disappeared by the final sampling date. The density of B. glandula was highest and peaked early in the mid zone (zone: p=0.003, zone × time: p=0.009, Fig. 4.2b). Space occupancy of B. glandula was also greatest in the mid zone, and increased over time, despite declining density, because early settling barnacles grew larger (Cover: zone: p<0.001, zone × time: p=0.006, Fig. 4.3b). Balanus glandula were less dense and lower in cover in warm treatments (Density & Cover: temperature: p<0.001, see Table 4.4 for full RM-ANOVA). Warming consistently reduced density over time (temperature × time: p=0.089) but the difference in cover between treatments increased; B. glandula cover stayed low on warm plates but increased on cool plates (temperature × time: p<0.001). Warming reduced density similarly in both zones (temperature × zone: p=0.287), but percent cover was more drastically reduced in the mid zone (temperature × zone: p=0.004), and this effect increased over the summer (temperature × zone × time: p=0.039). By March 2010, barnacles mostly consisted of adults, and there was no longer a significant effect of the temperature treatment on density (temperature: p=0.159, Table 4.4c), however percent cover remained lower in warm treatments (temperature: p=0.002, Table 4.4d). Warming had a similar effect on both the density and percent cover of C. dalli. Although the population of C. dalli was similar at both intertidal heights (Density & Cover: zone: p>0.350, see Table 4.5 for full RM-ANOVA), the timing of peak abundance was different between zones (Density & Cover: zone × time: p<0.001). Density peaked in May in the mid zone and in July in the high zone (Fig. 4.2c,d), while cover steadily increased over time and did so more quickly in the high zone (Fig. 4.3). The population of C. dalli was smaller and occupied less space in warm treatments (Density: temperature: p=0.032; Cover: temperature: p=0.042), and this effect was consistent in both zones and over time (Density & Cover: temperature × zone, temperature × time, temperature × zone × time: p>0.050). By March, there was still a significant effect of warming (Density: temperature: p=0.015; Cover: temp: p=0.009, Table 4.5c,d). This was largely due to differences in the high zone, where two cool plates that were sparsely populated with C. dalli were lost between the January and March sampling dates, causing the apparent mean density and cover of that treatment to increase, while the abundance and occupancy on warm plates stayed low.

70

Source df SS F P (a) Density – RM-ANOVA Between-subjects temp (t) 1,20 37.501 <0.001 zone (z) 1,20 11.637 0.003 t x z 1,20 1.196 0.287 Within-subjects time 6,15 14.145 <0.001 temp x time 6,15 2.299 0.089 zone x time 6,15 4.449 0.009 t x z x time 6,15 0.649 0.690 Mauchly criterion = 0.290, df = 20, P = 0.340

(b) Percent Cover – RM-ANOVA Between-subjects temp 1,20 64.305 <0.001 zone 1,20 33.338 <0.001 t x z 1,20 10.646 0.004 Within-subjects time 4.6,92.1 7.843 <0.001 temp x time 4.6,92.1 14.665 <0.001 zone x time 4.6,92.1 3.669 0.006 t x z x time 4.6,92.1 2.512 0.039 Mauchly criterion = 0.025, df = 20, P < 0.001

(c) Density – ANOVA (March 2010) temp 1 0.849 2.28 0.159 zone 1 0.892 2.401 0.149 t x z 1 0.001 0.001 0.980 error 11 4.085 total 14 6.679

(d) Percent Cover – ANOVA (March 2010) temp 1 1.765 16.056 0.002 zone 1 0.014 0.125 0.730 t x z 1 0.002 0.023 0.883 error 11 1.209 total 14 3.724

Table 4.4: Statistical results for Balanus glandula density and percent cover. Repeated Measures ANOVA for summer data for (a) density and (b) percent cover and ANOVA for (c) density and (d) percent cover data from the last time point. Degrees of freedom (df) listed as numerator, denominator for RM-ANOVA. Bolded values are significant at α = 0.05. Degrees of freedom & P-value adjusted by Huynh-Feldt ε in RM-ANOVA for Percent Cover.

71 Source df SS F P (a) Density – RM-ANOVA Between-subjects temp (t) 1,20 5.310 0.032 zone (z) 1,20 0.851 0.367 t x z 1,20 0.164 0.690 Within-subjects time 5.3,105.1 3.041 0.012 temp x time 5.3,105.1 2.161 0.061 zone x time 5.3,105.1 14.792 <0.001 t x z x time 5.3,105.1 1.195 0.316 Mauchly criterion = 0.072, df = 20, P < 0.001

(b) Percent Cover – RM-ANOVA Between-subjects temp 1,20 4.715 0.042 zone 1,20 0.010 0.921 t x z 1,20 0.041 0.842 Within-subjects time 18,120 5.348 <0.001 temp x time 18,120 1.446 0.203 zone x time 18,120 5.337 <0.001 t x z x time 18,120 1.374 0.231 Mauchly criterion = 0.144, df = 20, P = 0.023

(c) Density – ANOVA (March 2010) temp 1 2.903 8.268 0.015 zone 1 0.002 0.007 0.935 t x z 1 0.454 1.292 0.280 error 11 3.863 total 14 10.416

(d) Percent Cover – ANOVA (March 2010) temp 1 1.538 10.022 0.009 zone 1 0.004 0.025 0.877 t x z 1 0.154 1.002 0.338 error 11 1.688 total 14 4.850

Table 4.5: Statistical results for Chthamalus dalli density and percent cover. Repeated Measures ANOVA for summer data for (a) density and (b) percent cover and ANOVA for (c) density and (d) percent cover data from the last time point. Degrees of freedom (df) listed as numerator, denominator for RM-ANOVA. Bolded values are significant at α = 0.05. Degrees of freedom & P-value adjusted by Huynh-Feldt ε in RM-ANOVA for density and percent cover.

72 4.4.3 Barnacle survivorship

Barnacle survivorship was assessed in two ways: examining the relative abundance of empty tests and following young individuals through time to determine their survivorship. Dead barnacles (empty tests) were usually small and difficult to identify to species because all shells became white quickly, so data were pooled among species in analyses. The average proportion of dead barnacles was higher in warm treatments (temperature: p=0.019, Fig. 4.5, Table 4.6). There was also a higher proportion of dead barnacles during mid summer in the mid zone (June 24 – July 19), but the proportion of empty tests peaked during late summer (August 4 – September 15) in the high zone (zone × time: p=0.034). There was a trend towards a greater difference in dead barnacles between temperature treatments in the mid zone, although it was not significant (temperature × zone: p=0.061).

Figure 4.5: Mean (± s.e.) of the proportion of dead barnacles (empty tests) for B. glandula and C. dalli combined, in each of (a) high and (b) mid zones from April 2009 to March 2010.

73

Source df SS F P (a) RM-ANOVA (through August 2009) n=24 Between-subjects temp (t) 1,20 6.538 0.019 zone (z) 1,20 0.508 0.484 t x z 1,20 3.945 0.061 Within-subjects time 3.3,66.4 4.814 0.003 temp x time 3.3,66.4 1.806 0.149 zone x time 3.3,66.4 2.965 0.034 t x z x time 3.3,66.4 2.308 0.078 Mauchly criterion = 0.003, df =20, P < 0.001

(b) ANOVA (March 2010 only) n=15 temp 1 0.166 0.226 0.644 zone 1 3.609 4.914 0.048 t x z 1 0.166 0.226 0.644 error 11 8.080 total 14 13.929

Table 4.6: Proportion of dead barnacles estimated by empty tests for (a) summer and (b) the last sampling date. Species are pooled. RM-ANOVA conducted on raw (untransformed) data. ANOVA on last date conducted on log x +1 transformed data. Degrees of freedom & P-value adjusted by Huynh-Feldt ε in RM-ANOVA.

Survivorship of cyprids was significantly lower (1% in cool treatments; Fig. 4.6, Table 4.7) than survivorship for newly metamorphosed barnacles (57% in cool treatments) between June and August (life stage: p<0.001; species were pooled for analysis as cyprids could not be reliably identified from photos). Warming reduced survivorship of all young barnacles (temperature: p=0.040, Table 4.7), but more severely affected newly metamorphosed barnacles (temperature × life stage: p=0.048). However, the interactive effect was not strong because nearly all cyprids died regardless of treatment. Further, post-hoc comparisons revealed that the survivorship of cyprids was not significantly different between temperature treatments. From June to September, newly metamorphosed barnacle survivorship was reduced in warm treatments (Fig. 4.7), especially in the high zone, though the effect of temperature did not differ between barnacle species (temperature: p<0.001, temperature × zone: p=0.007, temperature × spp: p=0.324, see Table 4.8 for full RM-ANOVA). Overall, fewer B. glandula survived to September (spp: p=0.002), and B. glandula survivorship was more similar between zones compared to C. dalli survivorship (zone × spp: p<0.001; Table 4.8).

74

Figure 4.6: Total percent survivorship for cyprids and newly metamorphosed barnacles in cool and warm treatments. 25 barnacles per plate were followed over a two-month period from June 5 to August 4, 2009 in the high and mid zones and the proportion of surviving cyprids and newly metamorphosed barnacles per plate was determined. Data were pooled for B. glandula and C. dalli because identification of cyprids from photos was not possible. Species-specific data for newly metamorphosed barnacles can be found in Fig. 4.7. Averages and standard error limits are back transformed. Letters above bars represent a Tukey post-hoc comparison from the analysis in Table 4.7.

Source df SS F P temp (t) 1 6.226 4.600 0.040 zone (z) 1 4.758 3.516 0.070 life stage (ls) 1 50.180 37.077 < 0.001 t x z 1 0.280 0.207 0.652 t x ls 1 5.715 4.222 0.048 z x ls 1 1.071 0.791 0.380 t x z x ls 1 3.641 2.690 0.111 error 32 43.309 total 39 140.050

Table 4.7: Statistical results of a 3-way ANOVA, testing for the effect of the temperature treatment, intertidal zone, and life stage (cyprid or newly metamorphosed) on barnacle survivorship. Additional details provided in caption for Figure 4.6, above.

75

Figure 4.7: Survivorship of barnacles that recruited prior to June 5th. Average (± s.e.) total survivorship of newly metamorphosed (a,b) B. glandula and (c,d) C. dalli in cool and warm treatments in both zones (per 43.56 cm2 plate). 25 barnacles per plate were followed over a three- month period from June 5 to August 4 and August 4 to September 15, 2009. Initial assessment date (June 5) is included in graphs for visual interpretation but was not included in analyses.

Source df F P Between-subjects temperature (t) 1,26 29.460 <0.001 zone (z) 1,26 28.399 <0.001 species (spp) 1,26 12.319 0.002 t x z 1,26 8.553 0.007 t x spp 1,26 1.009 0.324 z x spp 1,26 17.428 <0.001 t x z x spp 1,26 0.450 0.508 Within-subjects time 1,26 18.723 <0.001 temp x time 1,26 9.625 0.005 zone x time 1,26 11.046 0.003 spp x time 1,26 3.772 0.063 t x z x time 1,26 10.199 0.004 t x spp x time 1,26 3.284 0.082 z x spp x time 1,26 4.134 0.052 t x z x spp x time 1,26 0.654 0.426

Table 4.8: Survivorship of newly metamorphosed barnacles. RM-ANOVA for selected summer dates (see methods). Data were arcsine transformed prior to analyses.

76 To determine which effects were most important for each species, I examined survivorship of barnacle species separately. Warming significantly decreased survivorship of newly metamorphosed B. glandula (temperature: p<0.001, Table 4.9a) and the difference between treatments was greater overall in the high zone because survivorship on cool plates was higher compared to the mid zone (temperature × zone: p=0.029, Fig. 4.7a). The effect of warming varied over the summer in each zone (temperature × zone × time: p=0.200). In the high zone, survivorship was reduced early and then stayed relatively constant within each temperature treatment. In the mid zone, survivorship declined to nearly zero in both treatments by the end of the summer, but did so more slowly on cool plates, which supports the high density of empty tests observed in that treatment over the summer (Fig. 4.5b). While collecting data for the growth rate analysis, I also observed that very few B. glandula survived for the entire two-month period (the minimum criteria for inclusion in that analysis) in warm high zone treatments (note the low sample sizes in summer warm treatments in Fig. 4.8). Despite frequent summer settlement (Fig. 4.2a,b), there was some thermal barrier prohibiting survivorship over the summer. A similar but more polarized pattern was observed for C. dalli. While warming reduced survivorship in both zones, the difference was greater in the high zone (temperature: p=0.018, temperature × zone: p=0.046, Table 4.9b, Fig 4.7c,d). This was due to almost 100% survivorship on cool plates and 41% survivorship on warm plates in the high zone compared to very low survivorship in both treatments in the mid zone. The effect of the temperature treatment was also constant for both sampling dates (all time interactions were non-significant, Table 4.9b).

77 Source df F P (a) Balanus glandula RM-ANOVA n=16 Between-subjects temp (t) 1,12 50.368 <0.001 zone (z) 1,12 1.620 0.227 t x z 1,12 6.186 0.029 Within-subjects time 1,12 17.468 0.004 temp x time 1,12 10.736 0.007 zone x time 1,12 12.755 0.004 t x z x time 1,12 7.120 0.020

(b) Chthamalus dalli RM-ANOVA n=18 Between-subjects temp 1,14 7.240 0.018 zone 1,14 33.416 <0.001 t x z 1,14 4.783 0.046 Within-subjects time 1,14 3.250 0.093 temp x time 1,14 0.951 0.346 zone x time 1,14 0.951 0.346 t x z x time 1,14 3.250 0.093

Table 4.9: Survivorship of newly metamorphosed (a) B. glandula and (b) C. dalli. RM-ANOVA for selected summer dates (see methods) for each species. Data were arcsine transformed prior to analyses.

4.4.4 Barnacle growth rate and size

Warming reduced growth rates in both species during summer months (temperature: p=0.016, Table 4.10, Fig. 4.8). On average, C. dalli growth rates were 64% slower than B. glandula in summer, and 58% slower in winter (Summer: spp: p<0.001; Winter: spp: p=0.006, Table 4.10). However, the effect of temperature did not differ between species in either season (Summer & Winter: temperature × spp: p>0.400). When analyzed alone, growth rates of juvenile B. glandula were not affected by temperature in either season (Summer & Winter: temperature: p>0.175, Table 4.11a). Low B. glandula survivorship on warm plates caused low sample size for the growth rate analysis, potentially preventing detection of a temperature effect. However, there was a trend towards slower growth rates in warm treatments in the summer (Fig. 4.8a, Table 4.11a).

78 Figure 4.8: Mean daily growth rates of juvenile (a,b) B. glandula and (c,d) C. dalli in cool and warm treatments (note the differences in y-axis scaling). The size trajectories of nine barnacles (per plate) that survived until the final sampling date were followed and averaged per plate. Growth rates were calculated from the two-month period following initial settlement. Summer (left) rates are from barnacles that settled in May-August. Winter (right) rates are for barnacles settling in September-January. Averages and standard error limits were back-transformed. Numbers on bars are replicate plates per treatment.

Source df SS F P (a) Summer, n=45 temperature (t) 1 1.060 x 10-2 6.371 0.016 zone (z) 1 1.984 x 10-3 1.192 0.282 species (spp) 1 5.670 x 10-2 34.071 <0.001 t x z 1 9.542 x 10-4 0.574 0.454 t x spp 1 8.959 x 10-4 0.538 0.468 z x spp 1 2.952 x 10-5 0.017 0.895 t x z x spp 1 1.586 x 10-3 0.953 0.335 error 37 6.156 x 10-2 total 44 1.566 x 10-1

(b) Winter, n=24 temperature 1 3.267 x 10-5 0.020 0.888 zone 1 5.868 x 10-3 3.687 0.073 spp 1 1.556 x 10-2 9.775 0.006 t x z 1 1.053 x 10-5 0.007 0.936 t x spp 1 2.611 x 10-5 0.016 0.899 z x spp 1 3.963 x 10-3 2.489 0.134 t x z x spp 1 3.487 x 10-3 0.022 0.885 error 16 2.547 x 10-2 total 23 4.950 x 10-2

Table 4.10: Statistical results for the effect of treatments on barnacle growth rates. ANOVA for (a) summer and (b) winter data. Data were log(x+1) transformed prior to analyses. Additional details provided in the caption for Figure 4.8, above.

79

Source df SS F P (a) Balanus glandula Summer, n=19 temperature (t) 1 6.99 x 10-3 1.983 0.179 zone (z) 1 9.90 x 10-4 0.280 0.604 t x z 1 1.98 x 10-3 0.562 0.465 error 15 5.29 x 10-2 total 18 6.54 x 10-2

(b) Balanus glandula Winter, n=12 temperature 1 5.15 x 10-5 0.020 0.892 zone 1 8.56 x 10-3 3.295 0.107 t x z 1 3.68 x 10-5 0.014 0.908 error 8 2.08 x 10-2 total 11 3.13 x 10-2

(c) Chthamalus dalli Summer, n=26 temperature 1 3.62 x 10-3 9.198 0.006 zone 1 1.04 x 10-3 2.638 0.119 t x z 1 5.42 x 10-5 0.138 0.714 error 22 8.65 x 10-3 total 25 1.34 x 10-2

(d) Chthamalus dalli Winter, n=12 temperature 1 2.10 x 10-7 <0.001 0.985 zone 1 1.08 x 10-4 0.185 0.679 t x z 1 4.10 x 10-6 0.007 0.935 error 8 4.68 x 10-3 total 11 4.80 x 10-3

Table 4.11: Statistical results (ANOVA) for the effect of treatments on (a,b) B. glandula and (c,d) C. dalli growth rates, in (a, c) summer and (b, d) winter. Data were log(x+1) transformed prior to analyses.

80 Overall, C. dalli growth rates were significantly reduced on warm plates in the summer (temperature: p=0.006, Table 4.11c), but there was no carry-over effect in the winter (Fig. 4.8d, Table 4.11d). Over the course of the experiment, C. dalli growth rates increased from May- September, then decreased in winter (Fig. 4.9a). This caused body size to increase exponentially during summer, then linearly from September-March (Fig. 4.9b). Growth rates were depressed on warm plates, especially in warmer summer months (temperature: p<0.001, temperature × time: p<0.015, Table 4.12). This caused average body sizes to diverge between treatments in mid summer (temperature: p<0.003, temperature × time: p<0.001, Table 4.13). To determine whether the temperature treatment affected growth rates and body size for C. dalli cohorts settling at different times of the year, I conducted two complementary analyses. It was not possible to perform comparable analyses for the B. glandula data set due to sample size constraints. First, I examined the effect of seasonal differences in settlement period, for barnacles of approximately the same age (37-48 days old; Table 4.12, Fig. 4.9 - solid grey boxes). For young C. dalli, the temporal pattern of growth rates was different between cohorts because July settlers experienced a steeper growth curve than earlier settlers (settlement date × time: p=0.006, Table 4.12a). Thus, at age 37-48 days, growth rates for July settlers were higher on average (0.082 mm2/day) than for earlier settlers (May: 0.046 mm2/day; June: 0.052 mm2/day), even though July settlers were slightly younger. Despite these differences in growth rate, barnacle size did not differ between cohorts over that period, and the relationship did not change over time (Table 4.13a). In addition, warming reduced the growth rates and body sizes of all cohorts similarly, and the effect did not change over time (temperature × settlement date: p>0.250, temperature × settlement date × time: p>0.250, Table 4.12a, Table 4.13a). In a second analysis, I examined the effect of different barnacle ages, over the same seasonal period (July-September; Fig. 4.9 - dashed grey boxes). I found that C. dalli were larger overall if they settled earlier (i.e. were older; settlement date: p<0.001, Table 4.13b), due to time since settlement and compounded by faster growth rates associated with larger, older individuals (Growth rate: settlement date: p=0.005, Table 4.12b). Over the same period, warming affected the growth rate of all cohorts similarly (temperature × settlement date: p=0.331, Table 4.12b), although warming affected the body size of cohorts differently (temperature × settlement date: p=0.019, Table 4.13b). Warming reduced the size of older (earlier settling) barnacles more than younger (later settling) barnacles. In fact, barnacles in warm treatments were similar in size on a

81 given date, regardless of age, while barnacles in cool treatments were larger if they were older. However, after one year, the ability to detect this interaction disappeared due to loss of replicates (temperature × settlement date: p=0.118, Table 4.13c). In winter (February and March), there was no difference in growth rates in warm and cool treatments. By March, barnacles of different ages did not differ in size (settlement date: p=0.143, Table 4.13c), but summer differences caused barnacles on warm plates to be 25% smaller than barnacles on cool plates, on average (temperature: p=0.021, Table 4.13c).

Figure 4.9: Juvenile C. dalli (a) mean daily growth rates and (b) mean basal area in cool and warm treatments over time, for barnacles settling in May, June, and July in the high zone. Means (± s.e.) are calculated from plate averages. Solid and dashed grey boxes indicate data included in RM-ANOVAs assessing the effects of season and cohort age, respectively (growth: Table 4.12, size: Table 4.13). The dashed horizontal black line in (b) indicates the approximate size when C. dalli reaches maturity (14 mm2 basal area; Southward and Southward 1967). Averages and standard error limits were back-transformed.

82

(a) Seasonal signal Cohort period analyzed age (same age) May June 5-July 9 48 June July 9-Aug 9 41 July July 19-Aug 15 37 Source df F P Between-subjects temperature (t) 1,27 18.620 <0.001 settlement date (sd) 2,27 2.157 0.135 t x sd 2,27 0.080 0.923 Within-subjects time 1.9,53.6 86.649 <0.001 t x time 1.9,53.6 8.890 <0.001 sd x time 3.9,53.6 4.089 0.006 t x sd x time 3.9,53.6 1.031 0.399 Mauchly criterion = 0.741, df =2, P = 0.020

(b) Age signal Cohort period analyzed age (same period) May July 19-Sept 15 116 June July 19-Sept 15 83 July July 19-Sept 15 68 Source df F P Between-subjects temperature 1,25 40.379 <0.001 settlement date 2,25 6.494 0.005 t x sd 2,25 1.155 0.331 Within-subjects time 2.8,69.1 88.506 <0.001 t x time 2.8,69.1 3.905 0.014 sd x time 5.5,69.1 1.411 0.227 t x sd x time 5.5,69.1 1.692 0.142 Mauchly criterion = 0.521, df = 5, P < 0.001

Table 4.12: Statistical analysis (RM-ANOVA) of the effects of temperature treatment and different settlement dates on C. dalli growth rates in the high zone during summer (see Fig. 4.9a). Also listed are data used for each analysis. Effects were examined for (a) a seasonal signal, using data for barnacles of approximately the same age (solid grey boxes in Fig. 4.9a) and, (b) an age signal, using growth over the same time period (dashed grey boxes in Fig. 4.9a). Average growth rates (per plate) were calculated for barnacle cohorts that settled just prior to May 22, June 24, and July 9, 2009. N = 3-7 plates / treatment. Degrees of freedom & P-value adjusted by Huynh-Feldt ε for within-subjects effects.

83

(a) Seasonal signal age Cohort period analyzed (same age) (days) May May 22-July 9 48 June June 24-Aug 4 41 July July 9-Aug 15 37 Source df F P Between-subjects temperature (t) 1,27 11.489 0.002 settlement date (sd) 2,27 0.081 0.922 t x sd 2,27 0.957 0.397 Within-subjects time 1.7,47.0 270.733 <0.001 t x time 1.7,47.0 24.390 <0.001 sd x time 3.5,47.0 1.270 0.296 t x sd x time 3.5,47.0 1.310 0.282 Mauchly criterion = 0.112, df =5, P < 0.001

(b) Age signal age Cohort period analyzed (same period) (days) May July 9-Sept 15 116 June July 9-Sept 15 83 July July 9-Sept 15 68 Source df F P Between-subjects temperature 1,25 28.450 <0.001 settlement date 2,25 30.712 <0.001 t x sd 2,25 4.653 0.019 Within-subjects time 2.7,68.6 680.417 <0.001 t x time 2.7,68.6 7.805 <0.001 sd x time 5.5,68.6 5.717 <0.001 t x sd x time 5.5,68.6 1.694 0.142 Mauchly criterion = 0.054, df =9, P < 0.001

(c) Adult size Cohort period analyzed age (after one year) (days) May Mar 24 306 June Mar 24 273 July Mar 24 258 Source df F P temperature 1 10.982 0.021 settlement date 2 2.947 0.143 t x sd 2 3.371 0.118

Table 4.13: Statistical analysis (RM-ANOVA) of the effects of temperature treatment and different settlement dates on C. dalli basal area (mm2) in the high zone during summer (see Fig. 4.9b). Also listed are data used for each analysis. Effects were examined for (a) a seasonal signal, using data for barnacles of approximately the same age (solid grey boxes in Fig. 4.9b) and, (b) an age signal, using growth over the same period (dashed grey boxes in Fig. 4.9b). Average body size (per plate) was calculated for barnacle cohorts that settled just prior to May 22, June 24, and July 9, 2009. N = 3-7 plates / treatment. Degrees of freedom & P-value adjusted by Huynh-Feldt ε for within-subjects effects in RM-ANOVA. (c) Statistical results (ANOVA) of temperature treatment and settlement dates on body size at the end of the experiment (March 2010), n = 11 plates.

84 4.5 Discussion

Although species are generally well adapted to their current thermal regime, climate warming may drive organismal body temperatures beyond important physiological thresholds, ultimately leading to detrimental effects on populations and communities (Vasseur et al. 2014). Integrating warming effects across ontogeny in a natural setting can elucidate mechanisms behind warming responses that cannot be perceived by studying a single life stage. This has rarely been attempted (though see Radchuk et al. 2013), and has rarely been quantified entirely from field-based manipulations. Doing so may provide better estimates of species’ responses to climate change.

4.5.1 Efficacy of experimental temperature manipulation

The substratum was successfully heated in situ for barnacle settlement by ~2°C when emersed during summer months, creating a treatment that reflects projected 21st century atmospheric warming for the region (IPCC 2013). Globally, air temperatures are expected to increase faster than water temperature (Sutton et al. 2007), a trend which has been observed in coastal British Columbia since 1950 (Wilson 2006). Since intertidal organisms experience thermal stress during emersion (Hofmann and Somero 1995), increasing air temperatures may be more detrimental than increasing water temperatures. Warmed plates reached higher temperatures and spent more time above important thermal thresholds than cool plates and the surrounding bedrock. In addition, black plates in the high zone stayed hot for several days in a row (i.e., there were many weeks with daily maxima between 30-42°C), during periods of prolonged emersion from spring and neap tides. When solar radiation was high, black plates were also hotter for more hours within a day. In the mid intertidal, black plates were “hot” (ADM ≥ 33°C) more often than white plates, but for a third as many days as comparable plates in the high intertidal. Although high temperatures were recorded in both zones, prolonged thermal stress was much greater higher on shore. The prolonged nature of this stress is important, as chronic aerial stress decreases the ability of intertidal organisms to recover between stressful events (Pincebourde et al. 2008). In addition, the higher variability characteristic of warm plates, is expected to be more important than increasing mean alone, especially for mid-altitude ectotherms (Vasseur et al. 2014). Overall, the thermal characteristics of the warmed plates offer a realistic comparison to future thermal

85 regimes, which are expected to feature longer, more frequent, more variable, and more intense high temperature events (IPCC 2013).

4.5.2 Lethal effects of warming

The survivorship of young barnacles was lower, the proportion of empty tests was higher, and the overall abundance of barnacles was lower following moderate warming of the substrata, indicating higher mortality in warmed treatments. For barnacles at the cyprid stage in early June, survivorship over the following two months (June-August) was extremely low (less than 2%) and did not differ between temperature treatments. This result contradicts my hypothesis that younger barnacles would be more sensitive to warming. However, Lathlean & Minchinton (2012) used a similar in situ heating method with Australian barnacles and also did not find an effect of warming on early survivorship. Previous work in the northeast Pacific has shown that survivorship of barnacles is 46% in the four days following settlement (Gosselin and Qian 1996). Thus, it is likely that most recruiting barnacles died very early in the two-month estimation period used in this study; if the temperature treatment affected cyprid survivorship, it was unlikely to have been detected two months later. Over the same period, warming reduced survivorship of newly metamorphosed barnacles by 87%. This 60 day period coincided with 25 days when the daily maximum temperature on warm plates exceeded 33°C (an important thermal threshold for B. glandula), with half as many hot days on cool plates. The survivorship of newly metamorphosed B. glandula was generally lower than that of C. dalli, however, temperature negatively affected the survivorship of the two species similarly. Substratum warming was also more detrimental higher in the intertidal for both species, where thermal stress was greater. Aerial warming affected juvenile B. glandula populations by killing young barnacles. Although B. glandula settled on warm plates, early recruitment was reduced by 59% and almost none survived long enough (2 months) to be used in the growth rate analysis, prior to those that settled in September. On cool plates, B. glandula survivorship was ~50%, which is consistent with estimates from previous work in the northeast Pacific (Gosselin and Qian 1996). Survivorship on cool plates was greater higher in the intertidal. This was likely due to a reduction in competition for space and/or interference by green macroalgae, which were common in the mid zone during the summer but rare or absent in the high zone until September

86 (Chapter 3). Green filamentous algae can compete with newly metamorphosed barnacles by smothering them (Barnes 1955). Previous work that measured B. glandula responses to aerial warming found that Hsp70 expression became maximal at 33°C, suggesting that this temperature may represent an important thermal threshold (Berger and Emlet 2007). However, there was no observed sign of irreversible protein damage when barnacles were exposed to temperatures up to 34°C for 8.5 hours (the highest temperature tested), indicating that B. glandula are well adapted to their current thermal regime (Berger and Emlet 2007). In the current study, the ADM substratum temperature was above 33°C for many more consecutive hours and days on warm plates. Although B. glandula may be able to withstand temperatures in excess of this threshold for eight continuous hours on a few consecutive days (on cool plates), it is likely that the increased periods and frequency of hot (33-37°C) and very hot (37-41°C) temperatures were too thermally stressful for barnacles, causing irreversible protein damage and reductions in survivorship. The survivorship of newly metamorphosed barnacles depended on the local habitat (intertidal zone), and this effect was more pronounced for C. dalli. In the high zone, C. dalli survivorship was cut in half by warming, although ~50% of juveniles still survived on warm plates after three months. In the mid zone, survivorship was less than 5%, regardless of treatment, likely due to negative biological interactions (as with B. glandula; see above for details). Barnacle mortality can either result directly from environmental stress (Gedan et al. 2011) or indirectly from predators such as the dogwhelk Nucella ostrina, the sea star Pisaster ochraceus, or dipterans in the genus Oedoparena (Harley and Lopez 2003, Harley and O’Riley 2011). Nucella ostrina were not observed on the plates during surveys, nor when just submerged (personal observation). Many mobile consumers feed when submerged rather than while exposed at low tide (Dahlhoff et al. 2002), so it is possible that predation occurred outside of survey periods. In fact, P. ochraceus were sometimes seen on plates shortly after submergence in the mid zone. However P. ochraceus feed by pulling the entire barnacle off the substratum leaving only basal plates behind, so it is unlikely that this species contributed to the empty test mortality patterns I observed. Oedoparena have higher barnacle infestation rates in cooler areas (Harley and Lopez 2003), which would lead to the opposite pattern of mortality from what was observed. In addition, nocturnal predators were not observed on plates during nighttime surveys in winter.

87 Lastly, N. ostrina and small P. ochraceus were found under plates with equal frequency among heating treatments (Chapter 3). Therefore, I am confident that the change in barnacle abundance in warm treatments was due to the thermal manipulation rather than predation.

4.5.3 Sublethal effects of warming

Although C. dalli had slower growth rates and reached a smaller adult size than B. glandula, warming had a similarly negative effect on the growth of both species, which is consistent with results from similar studies (Lathlean and Minchinton 2012). Warming reduced growth rates when both species were considered together, but the influence of warming on B. glandula growth disappeared when it was analyzed alone. There was a non-significant trend towards reduced growth rates for B. glandula in warm treatments, indicating that the effect of warming was either weaker or more difficult to detect in this species. Two possibilities may explain this result; either severe thermal stress selectively eliminated less robust, slower-growing individuals or sample size was too low to detect a difference. For surviving C. dalli, sublethal effects of warming were evident in 34-51% slower growth rates during summer, which resulted in 25% smaller body sizes after one year. In cool treatments in the summer, older (earlier settling) barnacles were larger than younger ones on a given date. However, the same phenomenon was not present in warm treatments where barnacles of all ages remained approximately equally small. Indeed, hotter temperatures kept barnacles as small as if they had settled months later. This finding contradicts my hypothesis that later settling barnacles would be more sensitive to thermal stress. May settlers experienced two months of thermal stress more than July settlers, allowing more time for detrimental effects. The ability to detect this interaction disappeared after one year, although the overarching detrimental effect of warming on adult size remained pronounced. Thermally-driven changes in body size distributions within a population can have important consequences for individual performance and for overall population growth rates. Body size for sessile invertebrates involves trade offs between reproductive ability, intra- and interspecific competition (crowding), susceptibility to predation, and physiological limits. Chthamalus dalli reach maturity around 14 mm2 basal area (Southward and Southward 1967). Smaller barnacles have lower reproductive outputs than larger barnacles, though the relationship is not strong for C. dalli (r2 < 0.30; Wethey 1984b). When crowded, the per capita biomass of C.

88 dalli is greater compared to uncrowded individuals; shell weights are greater although their shells are taller and thinner (Wethey 1984b) because neighbors provide some structural support (Bertness et al. 1998). Crowded individuals also have greater somatic tissue weights and produce more eggs (Wethey 1984b). Because warming slowed growth rates, C. dalli of all ages were smaller than 14 mm2 by the winter reproductive season, and thus were unlikely to reproduce (Fig. 4.9b, dashed black line). Slower growth rates probably also lowered the likelihood of crowding, which is beneficial to C. dalli fitness, and has been shown to increase survivorship in thermally stressful conditions for other barnacle species (Bertness et al. 1999a). Smaller barnacles may be at a disadvantage in other ways as well; for example, they are more vulnerable to bulldozing by limpets (Miller and Carefoot 1989). On the other hand, barnacles benefit from being smaller by avoiding predation by Nucella spp. and Oedoparena, which select for larger barnacles (Paine 1981, Harley and Lopez 2003). A detailed exploration of the potential indirect effects of warming on body size is beyond the scope of this study but, generally, has been the topic of recent reviews (e.g. Sheridan and Bickford 2011).

4.5.4 Population responses and shifts in dominance

These results demonstrate that warming was detrimental to coexisting barnacle species, but that the two species responded somewhat differently to increased thermal stress. The negative effects of summer warming were still apparent one year later in C. dalli density and percent cover, however only B. glandula cover still showed signs of stress from the previous summer (density was not affected). This discrepancy is likely due to interspecific differences in phenology; C. dalli reproduce once in early spring (Dayton 1971), whereas B. glandula release larvae twice (early spring and fall) or continuously brood throughout the year (Barnes and Barnes 1956). Thus the negative thermal effects on C. dalli cyprids and juveniles in the summer carried through to the adult population density (and percent cover) in March. By contrast, B. glandula larvae continued to settle after the thermally stressful summer period into the free space available on black plates, causing population densities to be similar in both treatments. Since this caused the proportion of large to small barnacles to be lower on black plates (RLK personal observation), percent cover was lower. As a result, the carry over effect of summer warming likely caused fewer B. glandula and C. dalli to reach reproductive size by winter, reducing the fitness of the populations.

89 Competitive dominance of one species over the other cannot be assessed because space did not become limiting (according to Dayton 1971, space becomes limiting when total barnacle cover exceeds 50%), however, there was an apparent temperature-induced change in relative space occupancy during the thermally stressful portion of the year, which varied among intertidal heights. In the mid zone, B. glandula were 500% more abundant than C. dalli in the cool treatment, but only 16% more abundant in the warm treatment. In the high zone, B. glandula were 16% more abundant than C. dalli in the cool treatment, and were half as abundant in the warm treatment in late summer. Both barnacle species live near their thermal maxima in the rocky intertidal zone, but warming affected their populations differently; B. glandula were more severely affected than C. dalli, especially lower in the intertidal. The higher sensitivity of B. glandula to thermal stress is consistent with its lower distribution relative to C. dalli on Salt Spring Island. Further, B. glandula has a lower thermal tolerance than C. dalli (Liao & Harley unpublished data), thus the former are likely to fare worse with increasing thermal stress. Despite different thermal tolerances, I did not observe any difference in rates of survivorship between species. Nor did I observe taxonomic differences in the effect of warming on growth rates. Thus, this effect was likely driven by the extreme differences in the mid intertidal where C. dalli density and survivorship were lower than B. glandula, likely due to biotic interactions with other species (e.g. green filamentous algae). This caused extremely low C. dalli percent cover in both mid zone temperature treatments. In the high intertidal, C. dalli density and survivorship were higher than that of B. glandula, so the interspecific asymmetry of the temperature effect was less pronounced. The apparent thermal sensitivity of the barnacles likely depended on both direct (temperature) and indirect (biotic interactions) effects, although only direct effects were controlled for in this study.

4.5.5 Response to a changing climate

As the climate changes, oceans will warm at a slower rate than the air (Sutton et al. 2007), but in the long term, both will be important to barnacle survival. Under future warmer conditions, B. glandula may experience down-shore shifts in distribution (Harley 2011), reduced or delayed reproduction (Barnes and Barnes 1956, Hines 1978), increased development rate that results in smaller cyprids (Pfeiffer-Hoyt and McManus 2005, Emlet 2006), and reduced juvenile and adult survivorship (Berger and Emlet 2007, this study). Further, B. glandula may not have any scope

90 to acclimate to future warming (Berger and Emlet 2007). The current study confirms that increased temperatures are detrimental to B. glandula populations, barring rapid adaptation or a down-shore shift in intertidal zonation patterns. However, given the steep genetic cline between B. glandula populations north and south of San Francisco and the homogeneity of northern populations (Sotka et al. 2004), the probability of warm-adapted genotypes arriving from the south appears to be remote. In addition, recent work suggests that the potential for down-shore shifts in sessile invertebrates may be constrained by interactions with other species (Harley 2011). The upper vertical limit of B. glandula populations is negatively related to substratum ADM temperature in the Salish Sea. Over the last 50 years, summer ADM temperatures in the region have risen by 3.40°C, lowering the upper limit by approximately 30 cm (Harley 2011). Over the same period, the upper limit of mobile predator foraging ranges have remained unchanged, resulting in truncated vertical distributions of sessile prey (Harley 2011). C. dalli will undergo a series of complex changes as the climate warms. Slowed larvae development rates (Miller et al. 1989) would delay settlement, which may lead to reduced reproductive output in the following year. Low-tide thermal stress will slow juvenile growth rates (this study) but increased feeding rates in warmer water at high tide (Southward and Southward 1967) may curtail the impairment. If a compensatory effect fails to arise, slower growth rates will lead to delayed maturity. Barnacle reproduction depends on reaching a mature size and releasing larvae when seasonal algal blooms are prevalent. Thus, the age of first reproduction may be delayed from year one to year two of a barnacle’s life. For an organism with a four to six year lifespan (Southward and Southward 1967), warming could significantly decrease lifetime reproductive output. The multi-generational effects of warming in a largely open population are difficult to know, and will depend on the juvenile and adult responses demonstrated here as well as how larvae respond and whether larval supply is limiting. This illustrates the necessity to integrate multiple responses at multiple life stages (in multiple habitats – water and air) into climate models in order to make meaningful predictions of the climate change effects. Models using information for a single life stage (e.g., adults) or a single vital rate would mistakenly forecast population responses. For example, the oldest study of these barnacles found that adult feeding rates increase with warmer water (Southward and Southward 1967); using these data alone would indicate that the species will not confront any direct risk from global warming. Coupled with data from other studies, models may predict that

91 the negative effects of warming on growth, survivorship, and fecudity are somewhat compensated for by increased feeding rates, at the population level.

4.5.6 Conclusions

In summary, I showed that a small amount of substratum warming was detrimental, even for two highly thermally tolerant intertidal barnacle species and even in areas that were not up against their vertical upper limits. Although warming affected the survivorship, growth rates, and population sizes of both species similarly, interspecific variation in phenology relative to periods of thermal stress allowed the B. glandula population to recover somewhat from heat induced summer mortality, whereas the C. dalli population remained depleted after one year. Nonetheless, increased thermal stress drastically reduced population sizes and slowed growth rates of both species, and likely shifted population structure towards smaller or younger individuals, reducing overall fitness. Barnacles, particularly larger bodied B. glandula, are an important food source for predators (whelks and sea stars) and an important habitat modifier for many other intertidal species. Thus, reduced barnacle body and population sizes may also indirectly negatively impact the barnacle-dependent community.

92 Chapter 5 Warming Modifies the Effect of Herbivory on Community Dynamics

5.1 Synopsis

As the Earth’s climate changes so too do its ecosystems, due to shifts in abundance, biodiversity and interaction strengths among their constituent species. Although increasing temperature will simultaneously affect many aspects of ecological communities, disentangling the abiotic and biotic contributions will allow for more accurate predictive models of how climate change will affect natural systems. Often, the importance of indirect effects of climate change, mediated by changing species interactions, will be greater – albeit less well understood – than direct effects in determining the community- and ecosystem-level outcomes of global climate change. To determine how warming modified the effect of species interactions on community dynamics, I manipulated temperature and herbivore access to settlement plates, factorially in a 16-month long field experiment in the rocky intertidal zone of Salt Spring Island, Canada. Warming and limpet herbivory influenced the abundance of some taxa independently (e.g., Ulva sp., Chthamalus dalli), however warming strengthened the indirect facilitative effect of limpets on the barnacle Balanus glandula and modified the negative effect of limpets on benthic diatoms. Warming reduced species diversity and evenness (especially in early succession), but there was no interactive effect between the temperature treatment and herbivory. Community structure and the successional trajectory of community structure over the duration of the experiment depended on the interaction between warming and herbivores. Warming and herbivores together increased abiotic and biotic pressure on dominant species, resulting in markedly different trajectories due to herbivore disturbance and high species turnover. Despite the stochastic nature of development, warmed communities with herbivores ultimately lost the variability created by herbivore- associated disturbances, resulting in highly similar assemblages between warm and cool treatments. These results illustrate how environmental change can alter species interactions, with cascading impacts on community dynamics.

93 5.2 Introduction

Temperatures across the planet are predicted to warm by 1.0-3.7°C over land and 0.6-2.0°C in surface oceans by the end of the century (IPCC 2013). This warming will take the form of both increasing mean temperatures and increasing frequency and severity of thermal extremes (Trenberth 2012, IPCC 2013). However, much is still unknown about the responses of whole communities to warming given the expected reshuffling of assemblages and likely changes in arrangement and strength of species interactions (Lurgi et al. 2012). Numerous studies have now catalogued a range of direct population-level impacts (Parmesan and Yohe 2003), but indirect effects mediated by interspecific interactions have received considerably less attention despite the growing awareness and repeated calls for their inclusion in empirical data and predictive climate models (Ings et al. 2009, Gilg et al. 2012). This is a critical gap in our knowledge because responses of complex, multispecies systems cannot be extrapolated from studies of single species in isolation. Community-level responses to warming are beginning to receive more attention, but most studies on the topic still focus on documentation of impacts rather than identification of mechanisms. Numerous studies have reported changes in the geographic distribution or phenology of populations (Parmesan and Yohe 2003, Sorte et al. 2010, Poloczanska et al. 2013), which can result in novel species interactions and community composition. Others have observed warming related changes directly to community structure. For example, periods of warming have been linked to increases in the abundance of warm-adapted species and decreases in the abundance of cold-adapted species (Barry et al. 1995); have caused mass die offs of habitat forming species, shifting dominance in community assemblages (Ostrander et al. 2000); and have shifted the balance of intertidal communities from primary producers to sessile invertebrates (Schiel et al. 2004). In cases like these, the structure of assemblages changed due to environmental warming, which undoubtedly also involved changing species interactions. However it is often difficult to understand the mechanisms responsible for changes in assemblages, as both abiotic variables (temperature) and biotic interactions are shifting simultaneously. Although the thermal influence on communities has been relatively well studied, we are still unable to predict the effects of warming on species assemblages. To do so, knowledge of

94 both direct (species-level) and indirect (interaction-mediated) influences on community dynamics is required. In many cases, the importance of indirect effects of climate change, mediated by changing species interactions, may be greater than direct effects of warming in determining the community- and ecosystem-level outcomes of global climate change. For example, warming strengthened competition between shrubs and forbs, altering community structure in a Rocky Mountain meadow (Harte and Shaw 1995) and shifted the competitive dominance of trees and shrubs, slowing succession of Mediterranean shrublands (Prieto et al. 2009). Warming has also been shown to modify the effect of trophic interactions (herbivory) on community structure (Post and Pedersen 2008, Eklöf et al. 2012), but the effect on higher level processes, such as diversity and stability has been inconsistent among studies (Werner and Matthiessen 2012, Wang et al. 2012, Post 2013, Williams et al. 2013). Further, almost nothing is known about how these factors might interact to affect community development (i.e., succession), which will be particularly important given the predicted increase in disturbances related to climate change (Easterling 2000, IPCC 2013). The marine intertidal is an ideal habitat for studies of warming because many intertidal organisms already live near their upper thermal tolerance limits (Hofmann and Somero 1995). This makes rocky shores ‘an early warning system’ for the impacts of climate warming (Helmuth et al. 2006b). In the intertidal, temperature drives physiological performance and contributes to patterns of species distribution and community structure (Somero 2002, Harley 2003, Steinbeck et al. 2005). Species interactions such as predation and competition are also known to be sensitive to temperature in rocky intertidal habitats (Sanford 1999, Pincebourde et al. 2008, Yamane and Gilman 2009). Grazers also have a strong influence on rocky intertidal community structure (Poore et al. 2012). In the northeast Pacific, limpets (Lottia spp.) primarily consume microalgae (including spores of macrophytes), benthic diatoms, and crusts (Nicotri 1977, Dethier and Duggins 1984, Steneck et al. 1991). Limpets increase resource (bare space) availability by reducing diatom and algal cover (Nicotri 1977), and by ‘bulldozing’ young sessile species such as barnacles (Dayton 1971). Limpets can also alter species interactions; for example, grazing frequency and intensity determined the competitive dominance of crustose algae (Steneck et al. 1991). To date, studies that simultaneously increase (rather than decrease) temperature and manipulate biotic variables (e.g., herbivory) in situ are rare, and have not been undertaken in the intertidal zone.

95 In this study, I used passively warmed settlement plates and herbivore exclusions in combination to determine how in situ warming directly (via physical stress) and indirectly (via altered interspecific interactions) affected populations and community dynamics. Manipulating both factors simultaneously provided the opportunity to uncover the mechanistic contribution of direct and indirect thermal effects to community processes. I hypothesized that warming alone would reduce barnacle and limpet populations, shifting community structure from a barnacle- dominated state towards depauperate substratum or occupancy by opportunistic species, as has been found previously (Chapter 3). Grazing alone would reduce microalgal abundance, freeing space for settlement of grazer-tolerant macroalgal taxa and sessile invertebrates. Further, I hypothesized that increasing thermal stress would weaken limpet grazing pressure (as predicted by consumer stress models), shifting the community away from barnacles and macroalgae and towards microalgae and diatoms. Overall, this would cause temporal and spatial patchiness from limpet grazing to be lower, changing pathways of succession.

5.3 Methods

5.3.1 Study system

I conducted a 16-month-long field experiment to determine the effects of warming and herbivory on intertidal community structure. The experiment was located in Ruckle Provincial Park (48.752° N, -123.367° W) on the southeast shore of Salt Spring Island, one of the southern Gulf Islands in British Columbia, Canada. Tides around the Gulf Islands are mixed semidiurnal and had a maximum tidal range of 3.6 m during 2011-2012. Lower low tides in summer months (March-September) coincided with the hottest portion of the day (10am-3pm). In addition, the Gulf Islands are protected from both oceanic swell and wind waves by Vancouver Island (a large barrier island directly west of the island chain; Demes et al. 2012), and fall within the rain shadow of Vancouver Island mountains. The archipelago that includes the Gulf Islands (Canada) and San Juan Islands (USA) is considered a ‘hot spot’ for intertidal temperatures relative to the outer west coast (Helmuth et al. 2002). These conditions make Salt Spring Island an ideal location for the study of warming on intertidal invertebrates. The experimental design consisted of a 2 x 2 factorial combination of black and white settlement plates and limpet-exclusions and controls, with eight replicates each. Plates of each

96 treatment combination were arranged in a blocked design for a total of 32 experimental units, and the order of treatments within each block was randomly assigned. All plates were installed on a gently sloping (mean of 16°) east facing bench in the ‘mid’ zone (1.4 m above Canadian chart datum, estimated using the lowest astronomical tide) of the rocky intertidal in 2011-2012. There was no difference in the orientation of the plates among treatments (Table 5.1). Limpet inclusions were also attempted but proved ineffective due to rapid rates of emigration; densities were nearly identical to limpet-exclusion treatments despite regular, manual reintroduction of limpets throughout the experiment. Accordingly, the limpet enclosure treatments were removed from analyses.

Source df SS F P Angle temperature (t) 1 13.385 0.453 0.505 limpet (l) 2 25.737 0.435 0.650 t x l 2 81.119 1.371 0.265 residual 43 1271.734 total 48 1393.918

Direction temperature 1 0.007 0.020 0.887 limpet 2 0.081 0.126 0.882 t x l 2 0.183 0.285 0.754 residual 43 13.862 total 48 14.137

Table 5.1: Angle and direction of plates installed in the rocky intertidal in Ruckle Provincial Park on Salt Spring Island.

97 5.3.2 Experimental warming treatment

I used black- and white-bordered settlement plates to assess the effect of warming on community dynamics. Settlement plates were made of black or white High-Density Polyethylene (HDPE) ‘puckboard’ plates (Fig. 5.1a). Each puckboard plate was 15.25 × 15.25 cm with a centered 6.9 × 6.9 cm area of white epoxy (Fig. 5.1a; for details, see Chapter 3). The temperature of the settlement surface was monitored using iButton temperature loggers (Thermochron DS1921G model, resolution ± 0.5°C; Maxim/Dallas Semiconductor Corporation, Dallas, Texas, USA), placed in a central recessed hole located under the epoxy surface between the two sandwiched plates, such that the sensor side of the logger was in direct contact with the epoxy. Two to four loggers per treatment (depending on logger malfunction) recorded the temperature of plates once every 60 minutes. Rock temperature was monitored using three iButtons embedded in Z-Spar (A-788 Splash Zone Compound), adjacent to plates containing loggers.

5.3.3 Limpet treatments

Copper paint or sheeting has been successfully used to manipulate grazer density in intertidal experiments and most effectively excludes limpets and chitons, while minimally excluding other mobile organisms like littorines (Johnson 1992, Harley 2006). Creating a physical barrier with copper sheeting is both more effective against grazers and avoids the risk of exposing organisms to harmful chemicals that are often found in paints (Johnson 1992). To exclude limpets from plate surfaces, I wrapped the bottom plate of each settlement plate unit with 5 mm thick copper sheeting (Fig. 5.1). T-nuts were hammered into the underside of corner holes of the bottom plate to keep the copper sheeting in place and to secure the top plate with screws. Copper was replaced every six to eight months to maintain limpet treatments. I did not use a partial copper control to examine the effect of copper because they often result in partial artifacts (Johnson 1992), and confuse interpretation of results. At a nearby location in Washington, Johnson (1992) found that direct contact with run-off from copper reduced algal growth by 20%. In the current study, copper was positioned below the organisms of interest (Fig. 5.1), preventing direct run-off when plates are uncovered at low tide. The reduction to algal growth from copper was likely much lower, between 0-20%. Nevertheless, Johnson (1992)

98 concluded that this reduction was more than compensated for by the protection from grazers. Thus the effect of grazers on algae is a conservative estimate when using this method.

Figure 5.1: Passive solar heated plate design with copper limpet-exclusions. (a) Schematic of plate with copper wrapping; (b) black exclusion plate (after 15 months) showing the efficacy of copper at excluding limpets (Lottia spp.); the most abundant herbivores in the mid intertidal on Salt Spring Island. (c) White exclusion plate after 13 months.

99 Limpets were by far the most abundant herbivore in the mid intertidal zone on Salt Spring Island (Fig. 5.1b), and the guild included several species; L. pelta, L. scutum, and L. paradigitalis. Limpet exclusions plus manual removals were 53% effective overall (40% on black and 66% on white plates), compared to controls. However, because limpets were manually removed from exclusions at each census, reducing limpet density to zero, this figure is a conservative estimate. More frequent manual removals (e.g. in summer) also increased effectiveness. To determine whether grazing pressure differed between treatments during high tide, when plates were the same temperature, I conducted a snorkel survey of plates on August 28, 2012. The survey was conducted on a flooding tide, when plates were 0.5-2 m under water. High-tide survey data were compared to low-tide densities recorded the same (hot and sunny) day. The experiment was installed on May 3, 2011, was monitored at two to four week intervals during summer months (March-September) and every three months during winter (September-March), and concluded on August 28, 2012. At each census date, densities of invertebrates and percent cover of all algal species on the plate surfaces were estimated using quadrats. Understory algal cover was estimated by moving the canopy layers aside, thus algal estimates sometimes exceeded 100% cover. Barnacles were often dense, and were therefore subsampled on plates (31% of epoxy area counted). A 3 mm area at the edge of the epoxy was excluded from counts to avoid edge effects, as some organisms were observed to prefer the topographic heterogeneity provided by the raised epoxy edge. I counted the number of mobile grazers (primarily limpets) on the epoxy plus individuals on the surrounding puckboard because limpet recruits (<5 mm) were often abundant around the epoxy edge and I reasoned that limpets on both surfaces would have access to organisms on the epoxy. The black and white puckboard plate borders were scrubbed clean of any diatoms and macroalgae at every sampling date to maintain temperature treatments. When limpets and other mobile consumers were found on the puckboard, they were removed prior to cleaning then replaced or were gently cleaned around. For clarity, I will use the term “control” only when referring to limpet controls (when limpets were allowed access to plates) and “exclusion” to refer to treatments where limpets were blocked from plates by copper borders. For temperature treatments, I will refer to white plates as “cool” and black plates as “warm”.

100 5.3.4 Statistical analyses

The average daily maximum temperature was calculated for each plate. Temperature differences between treatments only existed during daytime low tides in the summer and not at high tide or in winter when low tides were at night. Due to logger failure, few individual plates had complete temperature records for the duration of the experiment. This could potentially bias thermal comparisons among plates due to thermal differences among the time periods over which temperatures were recorded. To circumvent this bias, I calculated the residuals for each plate from the grand mean on each sampling date. The average residual across dates for each plate was then calculated, and these values were used in a two-way analysis of variance (ANOVA) to determine if temperature varied among plate colors or limpet treatments. To determine how temperature treatment and limpet exclusions affected species abundance and community metrics (e.g. richness), I used Repeated Measures Analysis of Variance (RM-ANOVA) to take time correlations across sampling dates into consideration. This type of analysis is suitable for experimental designs that take multiple samples over time from the same experimental unit (Quinn and Keough 2002), as was the case here. I analyzed different periods of time for each taxa depending on arrival of individuals on the plates. For example, Fucus distichus was not found on plates in any treatment until November, so I analyzed data from Nov. 11, 2011 through the end of the experiment. RM-ANOVA requires that data are normally distributed and that sphericity (equal variances of the differences between all combinations of the groups) is not violated. Data were transformed to meet these assumptions to the best possible extent. When several transformations yielded normally distributed data but resulted in violations to sphericity, the transformation that resulted in the most non-significant value for sphericity was chosen. For responses where sphericity was still violated following transformation, the degrees of freedom were altered according to the Mauchly test using Huynh- Feldt Epsilon (VonEnde 2001). The block effect was not significant in any analyses, so was pooled. Statistics were performed in JMP Pro 9.0.3 (SAS Institute). Results were confirmed with linear mixed effects models using the lmer function from the lme4 package in R (Bates et al. 2011), using temperature treatment, limpet treatment, and date as fixed factors and plate as a random factor. Significance was evaluated using a Satterhwaite approximation for the degrees of freedom in the lmerTest library (Kuznetsova et al. 2012). Since patterns of significance were similar between the two methods, I present the results for RM-ANOVA.

101 I estimated the population level effect of limpets on each taxa or functional group by comparing taxa abundances in the limpet-control and limpet-exclusion treatments. A limpet effect was confirmed by a significant result from the ‘limpet’ or ‘limpet × time’ terms in RM- ANOVA (described above). Further, I reasoned that limpets and warming together could affect species abundances in ways not expected by examining each effect alone; warming could affect limpets directly and thus alter the way that limpets interact with other species (i.e., there would be emergent effects, as predicted by metabolic theory and the consumer stress model). Emergent effects can be identified by a significant interaction term (e.g., temperature × limpet) in traditional ANOVA designs. However ANOVA designs test for an additive effect, which is inappropriate for most ecological measures (Sih et al. 1998). Multiplicative emergent effects can be detected by running an ANOVA on log transformed data (Sih et al. 1998). Thus I also analyzed taxa and community metrics using log(x+1) transformations, which rarely changed the significance of any responses (results with the original transformation are presented and changes in significance due to log transformations are noted). There are three broad categories of emergent effects that can result from interacting variables, traditionally used in terms of multiple abiotic stressors (Crain et al. 2008), but which can be adapted to understand how a factor (e.g., temperature) modifies species interactions. In the simple case of four treatments, with a control treatment (or reference state, since it need not be the literal control), treatment A, treatment B, and treatment A × B – if the response to A × B is the product of the response to A and the response to B, the interaction is said to be multiplicative. The relative magnitude of the responses in the control, A only, and B only treatments, dictate how the response to the A × B treatment is interpreted (see Fig. 5.2 for hypothetical results using the current experimental design). When the observed A × B response deviates from the expected multiplicative response, the effect of A is ‘strengthened’ by B if the response is farther from the response in the control treatment, relative to the expected multiplicative treatment (Fig. 5.2). The effect of A is ‘weakened’ by B if the actual response falls in between that of the multiplicative and B, and the effect of A is ‘reversed’ by B when the actual response falls on the opposite side of the response to B alone, as compared to the multiplicative case (Fig. 5.2). When the interaction term in the RM-ANOVA was significant (using the log transformation), I investigated the presence of these emergent effects for all taxa and diversity metrics. To investigate whether the copper treatment and warming together had unexpected

102 consequences (i.e., emergent effects), I used the cool limpet-control as the reference state. I then evaluated whether warming modified the effect of the copper exclusion on limpet abundance by comparing the warm, limpet-exclusion treatment to the null multiplicative model. For all other responses, I was interested in how warming modified the effect of limpet presence (either through direct herbivory or some indirect mechanism) on taxa and diversity metrics. Thus, I used the cool limpet-exclusion treatment as the reference state, and compared the response in the warm, limpet-control treatment to a null multiplicative model (modeled after the hypothetical examples in Fig. 5.2). Since my experimental design involved repeated measures, I examined whether the temperature × limpet or the temperature × limpet × time effects were significant. If the former, I calculated averages of the response across sampling dates to compare to the multiplicative null. If the latter was significant, the observed (A × B) treatment and null multiplicative model were compared for each date (relative to the reference state). Significant differences from the null were determined by comparing overlap of 95% confidence intervals.

103

104 Figure 5.2: Conceptual approach to interpreting the influence of warming on the effect of limpets on a focal taxa (given here in percent cover – numbers at the bottom of bars). Four possible scenarios are illustrated using the current experimental design, where Con: control (cool, limpet-exclusion; hatched blue bars), Limp: effect of limpets only (cool, limpet-control; solid blue bars), Warm: effect of warming only (warm, limpet-exclusion; hatched red bars). The remaining four bars represent possible interactive effects when warming and limpets are combined (Limp × Warm). The response may simply be the product of the two treatments (Mult: multiplicative; grey bars), which serves as the null model. Whether warming strengthens (Str), weakens (Weak), or reverses (Rev) the effect of limpets (solid red bars) depends on the response relative to the warming only treatment (dashed horizontal line) and the multiplicative null expectation (solid horizontal line). Compared to the control, there are four possible outcomes for the effect of limpets only and warming only; (a) both may affect the response positively, (b) limpets may affect the response positively, while warming affects the response negatively, (c) the reverse may be true, or (d) both variables may affect the response negatively. Although the relative magnitudes of responses appear different in each scenario, warming strengthens the effect of limpets when the response is farther from the control than the multiplicative case, warming weakens the effect of limpets when the response falls between the warming only and multiplicative cases. The effect of limpets is reversed by warming when the sign (+ or -) of (Limp – Con) differs from the sign of ((Limp × Warm) – Warm); in other words, when the response value falls on the opposite side of the dashed line from the strengthened and weakened cases.

To test my hypothesis that community structure would be affected by temperature and limpets, I analyzed the assemblage structure with multivariate statistics, using PRIMER 6 software (Clarke and Warwick 2001). I converted density estimates of invertebrates into percent cover estimates by using the average of organism sizes (organism area was measured from photographs of plates). I excluded all limpet species from these analyses since they were manipulated as part of the design. I investigated the effect of treatments on community structure from three dates that represented distinct periods in the timeline of the experiment (see Results); July 27, 2011, April 9, 2012, and August 28, 2012. Community structure was estimated by calculating the Bray-Curtis similarity matrix of the log(x+1) transformed percent cover of each species. I used a log(x+1) transformation to downweight the contribution of quantitatively dominant species to the similarities calculated between samples (Clarke and Gorley 2006) and maintain consistency in transformations for ease of interpretation with other taxonomic analyses. Variation in community assemblage was compared using permutational multivariate analysis of variance (PERMANOVA) with temperature and limpet treatments as fixed factors and type III sums of squares. Each term in the analysis was tested using 9999 random permutations of the

105 appropriate units. Variability in community structure was examined with permutational analysis of multivariate dispersions (PERMDISP) which tests for the homogeneity in dispersion, essentially a multivariate version of Levene’s test (Anderson et al. 2008). When a significant interaction (temperature × limpet) was detected in PERMANOVA or PERMDISP, pair-wise comparisons were also calculated. Non-metric multidimensional scalings (nMDS), based on a Bray-Curtis similarity matrix, were used to visualize partitioning in assemblage structure. In addition, a SIMPER analysis was conducted using treatments where significant effects were identified to determine which species were the principal contributors to dissimilarity in community structure. Differences in successional trajectories (the pattern of community change through time) were compared with a second stage analysis. For each of the 32 replicates (in four treatments), I created separate resemblance matrices that each included all sampling dates, based on log(x+1) transformed data and a Bray-Curtis similarity. Spearman correlations were calculated between the 32 matrices, which generated a second stage resemblance matrix. This second stage matrix is essentially the similarity between replicate recovery trends regardless of the identity of taxa (Clarke et al. 2006), and when visualized using nMDS, is used to create a point that represents the entire recovery trajectory of a single plate. I then used PERMANOVA to determine if trajectories differed between treatments, independent of the identities of functional groups. I also calculated pairwise comparisons among all treatments. To identify differences in trajectory variability, I also conducted a second stage PERMDISP with pairwise comparisons. This was done by calculating the dissimilarities among trajectories and running a principal coordinates analysis on the dissimilarity matrix. I visualized these trajectories in two ways; first, I plotted the entire path of the trajectory for each treatment by generating the nMDS coordinates (as above) for each replicate plate at each sampling date, then calculated the average nMDS coordinates of the eight replicates within each treatment combination. Second, to visualize the recovery variability within and between treatments, I also created an nMDS ordination of the second stage analysis, which displays a single point for the entire trajectory of each replicate plate. Points that are close together indicate similar trajectory routes, while distant points indicate that succession on plates took different routes.

106 5.4 Results

5.4.1 Treatment effectiveness

The average daily maximum (ADM) temperature (averaged across plates) varied seasonally and through the fortnightly tidal cycle. Sea surface temperatures in the intertidal varied from 14°C in summer to 7°C in winter. During low tides, intertidal substratum temperatures varied from 25- 42°C in summer, when low tides were during the day, to just below freezing in winter, when low tides were at night (September-March). Plates reached higher temperatures (ADM 25-42°C) during spring tides, which coincided with the hottest portion of the day (10-3pm), and lower temperatures (ADM 10-20°C) during neap tides when daylight emersion time was less than four hours per day. Plates only differed in temperature when it was warm, and not when it was cool (winter, at night, when submerged during high tide). Once exposed by the receding tide on a sunny day, plates warmed by more than 20°C and the temperature on black and white plates diverged. Residual ADM plate temperatures were not different between limpet-exclusions and controls (Table 5.2a) so limpet treatments were pooled (Table 5.2b, Fig. 5.3). Residual ADM temperatures were consistently higher on black plates (residual ADM 0.72; ADM 19.02°C) compared to white plates (residual ADM -0.74; ADM 17.47°C) during summer months (temp

F2,21=5.991, p=0.009). The difference in ADM temperatures between black and white plates was 1.6°C, and the difference in the average absolute maximum on black plates was 6.0°C warmer than the average absolute maximum on white plates (Table 5.3). The variance in summer ADM temperature was higher on warm plates than cool plates (temperature F1,17=11.123, p=0.004) although there was no difference in the 10-90% or 25-75% range of temperature between plates. Plate temperatures encompassed the range of temperatures found on the nearby rock (Fig. 5.3), but the temperature of warm plates often exceeded that of the rock (Fig. 5.4).

107

Source df SS F P (a) Model including limpet treatment temperature (t) 1 7.647 16.402 0.001 limpet (l) 1 0.841 1.805 0.199 t x l 1 0.195 0.419 0.527 residual 15 6.993 total 18 18.165

(b) Model pooling limpet treatment, including Rock temp temperature 2 10.219 5.991 0.009 residual 21 17.910 total 23 28.129

Table 5.2: ANOVA table for ADM temperature residuals. (a) Comparing limpet treatments. (b) Pooling limpet treatments, comparing plate temperatures to the nearby rocky bench.

Figure 5.3: Residual ADM temperature in each temperature treatment compared to nearby rock temperatures. Control and limpet-exclusions were pooled within each temperature since they did not differ in temperature (corresponding statistical analysis can be found in Table 5.2b).

108 Temp Mean Mean Mean Mean Mean Mean trmt min (raw) (ADM) (residual) max variance Black -2.00 11.54 19.02 0.72 40.50 19.27 White -1.00 11.58 17.47 -0.74 34.50 13.82 Rock 8.50 12.96 21.18 -0.13 37.00 17.64

Table 5.3: Summary statistics for the temperature (°C) of plates and the nearby rock on Salt Spring Island, during the summers of 2011-2012 (when low tides were during the day). Mean min. and mean max. values represent the minimum and maximum values ever recorded on each plate or rock, averaged across replicates. Mean (raw) represents the mean temperature recorded on each plate or rock over two summers, averaged across replicates. Mean (ADM) was calculated by averaging the daily maximum temperature for each plate or rock over the experimental period. These means were then averaged across replicates within each treatment. Mean (residual) was calculated by averaging plate or rock residuals (calculated using the grand mean for each day) across time, then across replicates. The variance in temperature over two summers was calculated and averaged across replicates.

Figure 5.4: Daily maximum temperature, averaged across replicate iButtons, in each treatment (including the nearby rock) over 16 months.

109 Limpets (primarily L. pelta, L. scutum, and L. paradigitalis) were approximately twice as abundant on plates without copper than copper exclusions (Fig. 5.1b, Fig. 5.5a, Table 5.4). The density of limpets varied significantly over time. On plates where limpets had access, limpets did not appear in surveys until one month following plate installation and the density stayed low until winter (Fig 5.5a). Following the winter peak, density declined in spring 2012 before increasing again in late summer. Limpets were less abundant in warm treatments (temperature: p<0.001) and the difference between limpet-controls and exclusions varied between temperature treatments over time (temperature × limpet × time: p=0.028, Table 5.4). Results from the snorkel survey indicated that limpets were slightly more abundant during the low-tide survey than the high-tide survey (Fig. 5.6, Table 5.5). Limpets were also more abundant on cool and limpet- control plates (temperature: p<0.001; limpet: p<0.001), but the magnitude of these differences varied depending on whether plates were covered by water or not; warm plates reduced densities more during low tide, when thermal stress was highest (temperature × emersion: p=0.004). Since analyses revealed that warming modified the effect of the copper exclusion on limpet abundance and that this relationship changed over time (Table 5.4), possible emergent effects of warming and copper exclusions were explored. On each sampling date, I compared the warm exclusion to the null multiplicative model (using log transformed data), with the cool limpet-controls as the reference condition. Generally, exclusions alone and warming alone each reduced limpet densities, and the combination of warming and exclusion further reduced densities (Fig. 5.5b). The observed combined effect (warm exclusion) was indistinguishable from the null multiplicative expectation on 67% (9 out of 13) of sampling dates (Fig. 5.5c). In November, February, and August, warming and excluding with copper reduced limpet density more than expected based on the multiplicative model. In May 2012, the opposite was true, although this may have been an artifact of very low densities in all treatments (0-1 limpet / 100 cm2 in any treatment). Copper exclusions did not appear to exclude other grazers such as Littorina spp. or amphipods. The only statistically significant influence on non-limpet grazers was the limpet × time interaction (p=0.020, Table 5.4), which suggests that the differences between treatments fluctuated through time. There were no strong trends within these fluctuations, although non-limpet grazer densities tended to be greater in limpet-exclusions when treatment differences existed (Fig. D.1a). This pattern was consistent when Littorina spp. and amphipod densities were analyzed separately (data not shown).

110

Figure 5.5: Density of limpets over the duration of the experiment. (a) Average limpet density / 100 mm2 (± SE) in each treatment over 16 months. (b) The limpet density in the warm limpet- exclusion treatment (hashed red) was compared to a multiplicative null model (grey), calculated from the other three treatments (cool -, cool +, warm +). Density was averaged across all sampling dates for each treatment. Part b must be interpreted in light of part c because analyses revealed that the temperature x limpet x time effect was significant, however the temperature x limpet effect was not significant. (c) For each sampling date, the density in the warm limpet- exclusion treatment (dashed red) was compared to the expected abundance (grey), with reference to the cool limpet-controls (represented by the horizontal x-axis). Significant differences between observed and expected densities are shown with asterisks (determined by overlap of 95% CI). Values are back transformed.

111

sp. limpet

- Limpets Non grazers Predators glandula B. (barnacle) dalli C. (barnacle) Diatoms Ulva Green filaments Reds Crusts distichus F. algae Total Species Richness Diversity Evenness Time <0.001 <0.001 0.017 <0.001 <0.001 <0.001 <0.001 <0.001 0.007 <0.001 * 0.157 <0.001 <0.001 <0.001 0.022

Temperature (T) <0.001 0.130 0.947 <0.001 0.010 0.335 * 0.019 <0.001 0.033 0.414 0.539 0.633 0.407 0.043 0.011 T x time 0.002 0.089 0.193 <0.001 0.001 <0.001 0.001 <0.001 0.003 0.314 0.792 <0.001 0.581 0.001 0.001

Limpet (L) <0.001 0.195 0.036 <0.001 <0.001 <0.001 <0.001 0.086 0.020 0.021 0.842 <0.001 0.717 0.857 0.720 L x time <0.001 0.020 0.328 0.019 <0.001 0.112 <0.001 0.067 0.024 0.015 0.713 0.053 0.064 0.002 0.061

T x L 0.202 0.366 0.551 0.003 0.269 0.103 0.211 0.515 0.112 0.663 0.313 0.756 0.053 0.053 0.932 T x L x time 0.028 0.243 0.360 0.028 0.868 <0.001 0.093 0.618 0.016 0.956 0.114 * 0.049 0.311 0.231 0.688

Transformation log log sqt log log arcsine arcsine arcsine arcsine arcsine arcsine sqt log log raw Dates used 5/31 6/27 7/27 all all all 5/31 5/31 7/27 7/27 11/22 all all all all Sphericity <0.001 <0.001 cnc <0.001 <0.001 0.008 <0.001 cnc <0.001 cnc <0.001 0.177 0.002 0.002 0.442

Table 5.4: Repeated measures ANOVA p-values for key taxa and diversity metrics. Log, square root, and arcsine square root transformations were explored for all data sets. The transformation yielding the most normally distributed data and least sphericity is listed. When the p-value for Mauchly’s sphericity criterion was less than 0.05, the H-F epsilon corrected p-values were reported. ‘Dates used’ indicates the first date the taxon appeared in any treatment in the experiment and was the starting date used in analyses. Block and block × time were pooled because p > 0.250. Bolded p-values indicate significant effects at α = 0.05. n=8 for each treatment. Cnc = could not be calculated. Colors correspond to the general direction of the trend (also see Figures 5.5a, 5.9, D.1): for temperature effects, blue indicates higher abundance in cool treatments, red indicates higher abundance in warm treatments; for limpet effects, yellow indicates higher abundance in limpet-controls, green indicates higher abundance in limpet-exclusions; for the main effects × time: colors are same as the corresponding main effect when that pattern remained relatively consistent over time. When a switch occurred, the cell is grey. Asterisks (*) indicate a change in the significance of an effect when a log(x+1) transformation was used (for interpretation of multiplicative effects of temperature and limpets, see Fig. 5.8).

112

Figure 5.6: Limpet density (on epoxy and HDPE border) in each treatment during low and high- tide surveys on August 28, 2012. Letters that are different than each other indicate significant differences at α = 0.05 (statistical results are in Table 5.5). Hashed bars are densities during surveys at low tide, solid bars are densities during surveys at high tide.

Source df SS F P temp (t) 1 10.855 33.924 < 0.001 limpet (l) 1 88.479 279.503 < 0.001 emersion (e) 1 1.461 4.567 0.036 t x l 1 0.267 0.835 0.364 t x e 1 2.892 9.037 0.004 l x e 1 2.462 7.693 0.007 t x l x e 1 0.465 1.452 0.232 residual 84 26.879 total 91 134.732

Table 5.5: ANOVA table for limpet abundance during low-tide and high-tide surveys on August 28-29, 2012. Data were log(x+1) transformed to meet assumptions.

113 5.4.2 Functional groups

Succession and community structure differed among temperature and herbivory treatments. On cool limpet-control plates, diatoms and B. glandula dominated the early stages of succession (Fig. 5.7a, Fig. D.1). Other algal species (green filamentous algae, Ulva sp., crusts) and C. dalli barnacles were present at relatively low abundance through the winter and into the following summer. In May-June 2012, B. glandula settled again in high densities. Total percent cover of algae was highest in summer 2011 (when limpet abundance was lowest) and steadily declined to below 30% by the end of the experiment (Fig. D.1h). On cool plates where limpets were excluded, B. glandula were approximately 40% less abundant in summer 2011, compared to cool limpet-controls. Green filamentous algae, C. dalli, barnacle predators (e.g. Emplectonema gracile, Oedoparena sp. larvae), and other grazers (Littorina spp., amphipods) were present in low abundances in the first year of the experiment (Fig. 5.7b, Fig. D.1). Colonial benthic diatoms were the largest space occupier initially, and dominated plates through summer 2011 until September. In September, Ulva sp. began to colonize, and together with diatoms, was the major space occupier through winter. In spring, diatoms declined as Ulva cover increased to 80-90% through summer 2012. Balanus glandula recruited to plates in May-June 2012, but were 75% less abundant compared to cool controls at that time. Total algal percent cover exceeded 100% one month after the start of the experiment, and remained above 70% for the duration of the experiment (Fig. D.1h). Warm limpet-control treatments were characterized by several consecutive peaks in abundance (Fig. 5.7c). In summer 2011, barnacles, green filamentous algae, and diatoms peaked briefly and in rapid succession before declining sharply. In winter, Ulva, red algae (mostly Pyropia sp; Fig. D.1i), and diatoms displayed small peaks (30-60% cover) in abundance and declined by the following survey. In May-June 2012, about half as many B. glandula recruited to plates compared to cool limpet-controls, though the pattern of decline through the summer was similar. Diatoms, and to a lesser extent, Ulva and crusts (Fig. D.1j), were present on plates through the winter and summer 2012. Barnacle predators and non-limpet grazers were rare in any treatment at any time, but were more abundant in summer 2012 (Fig. D.1). Although total algal percent cover was high in the summer 2011 (~ 75%; when limpet densities were lowest), cover declined to 30% beginning in summer 2012 (despite low limpet abundance; Fig. D.1h). In general, warm control plates were more sparsely occupied than other treatments.

114 Figure 5.7: Mean abundance (+/- SE) of key taxa through time in each treatment; (a) cool, limpet-control, (b) cool, limpet-exclusion, (c) warm, limpet-control, (d) warm, limpet-exclusion. All species contributing to the similarities of community composition (Table 5.9) are displayed. Barnacle (B. glandula and C. dalli) density is displayed on the primary y-axis, percent cover of algal species (green filaments, Ulva sp., diatoms) is displayed on the secondary y-axis. For graphs of each taxa with all treatments overlaid, see Figure D.1.

115 Like warm controls, warm plates where limpets were excluded exhibited several brief peaks in abundance in summer 2011, although the peaks in barnacles were lower and the peaks in algae were higher (Fig. 5.7d). Chthamalus dalli and B. glandula initially recruited in very low densities and nearly disappeared by June 2011. Green filamentous algae appeared on plates in late May, reached 95% cover in June, then declined to zero by September 2011. A diatom mat formed (95-100% cover) as green filamentous algae started to decline in July, and diatoms remained relatively abundant through winter (35-70% cover). Ulva colonized in September and became abundant by April, and along with diatoms, was the major space occupier on plates through summer 2012. Total algal cover exceeded 100% (largely due to diatoms and Ulva) by June 2011 and remained high until the end of the experiment (Fig. D.1h).

5.4.3 Species interactions

Warming and herbivory affected most taxa independently (e.g. Ulva, C. dalli; Table 5.4), however warming modified the effect of limpets (T × L and/or T × L × time was significant) on the barnacle B. glandula and on diatoms. The interactions were examined in the context of ‘emergent effects’ to estimate whether warming strengthened or weakened the interaction more or less than expected. The response of B. glandula or diatoms in warm controls (+ limpets) was compared to a null multiplicative model, using cool exclusions (- limpets) as the reference condition. Using this treatment as the reference state (though it is not, strictly speaking, the control) allowed us to estimate the effect of limpet presence. RM-ANOVA of B. glandula abundance indicated that warming modified the effect of herbivore presence (T × L, p = 0.003) and that this relationship changed over time (T × L × time, p = 0.028). Averaged across dates, limpet presence positively affected the density of B. glandula, while warming negatively affected density. However, the combined effect was different than expected (Fig. 5.8a); warming strengthened the positive effect of limpet presence on B. glandula, or conversely, limpet presence weakened the negative effect of warming on B. glandula. Over time, the observed abundance differed from the expected value on 75% of sampling dates (Fig. 5.8b). On eight out of those nine dates, the relative pattern of treatments approximated that of the averaged values presented in Fig. 5.8a. On the last date, warming reversed the effect of limpets because limpets slightly decreased barnacle abundance in cool treatments and increased abundance in warm treatments.

116

Figure 5.8: (Left) Abundance of B. glandula and diatoms in each treatment, averaged across sampling dates. To determine whether warming strengthened (S), weakened (W), or reversed (R) the effect of herbivory, the abundance in the warm limpet-control treatment (solid red) was compared to a multiplicative null model (grey), calculated from the other three treatments; cool excl (hashed blue), cool con (solid blue), and warm excl (hashed red). (Right) For each sampling date, the warmed limpet-control (solid red) was compared to the expected abundance (grey). Both the observed and expected values are scaled to the abundance in the reference condition (cool limpet-exclusion), represented by the horizontal axis. The nature of the interaction is denoted when the observed abundance was significantly different from the expected (using 95% CI), and depended on the relative abundances in the treatments not shown (cool +, warm -). For example, on most sampling dates, B. glandula were more abundant in cool + compared to cool -, and were more abundant in warm + than warm -, although the difference was greater in the warm treatments, thus warming strengthened (S) the effect of the limpet treatment. In August 2012, B. glandula were slightly more abundant in cool – compared to cool +, though were much more abundant in warm con compared to warm -, thus warming reversed (R) the effect of the limpet treatment. For B. glandula, the (a) T × L and (b) T × L × time terms were significant in RM- ANOVA (Table 5.4). For diatoms (including Navicula), only the (d) T × L × time term was significant, thus (c) must be interpreted in light of (d).

117 Averaged over the entire experiment, diatoms were negatively affected by limpet presence and were positively affected by warming within each limpet treatment (Fig. 5.8c). However, these patterns were not consistent through time. The significant interaction with time

(T × L × time: F12,252=3.262, p<0.001) resulted from a ten week delay in establishment of diatoms on warm plates, followed by low cover in cool controls and higher but similar cover in the other treatments through mid summer 2012 (Fig. D.1e). In the last six weeks of the experiment, the amount of diatoms in treatments diverged, with the highest cover in warm exclusions. On half of sampling dates, the interactive effect of warming and herbivory was not different than expected by the null – in other words, warming did not change the strength of herbivory on these dates (Fig. 5.8d). On three dates (May 2011, Sept. 2011, and Aug 2012), diatoms were more abundant in the presence of limpets on cool plates, but this positive effect was reversed on warm plates. In November and February, warming weakened the negative effect of limpets on diatoms and in July 2012, warming strengthened the negative effect of limpets.

5.4.4 Diversity

Species richness, diversity, and evenness varied throughout the experiment (Fig. 5.9, Table 5.4). Species richness was not strongly affected by the main effects of temperature or limpet treatment (temperature: p=0.407; limpet: p=0.717). There was a trend towards lower richness in warm exclusions though it was not significant because richness in treatments became more similar in summer 2012 (temperature × limpet: p=0.053, Fig. 5.9a). Diversity displayed more differentiation between treatments (Fig. 5.9b, Table 5.4). Warming and the limpet exclusion independently reduced diversity through the first year (temperature: p=0.043; temperature × time: p=0.001; limpet × time: p=0.002), but, as with richness, treatments became more similar starting in the summer 2012. Evenness was similar in limpet treatments but warming significantly reduced evenness (temperature: p=0.011), and this difference was greatest in the first summer (temperature × time: p=0.001; Fig. 5.9c).

118

Figure 5.9: Average (a) species richness, (b) diversity, and (c) evenness over the duration of the experiment. Limpets were excluded from all metrics. Diversity was calculated with relative abundance data (compared to the maximum for each species), using the Shannon-wiener index. Red lines indicate diversity on warm plates, while blue indicates diversity on cool plates. Solid circles and solid lines represent limpet-control treatments while open squares and dashed lines indicate limpet-exclusions.

119 5.4.5 Seasonal community structure

Because of thermal seasonality and varying limpet abundances, there were distinct periods in the timeline of the experiment when each independent variable was exerting more or less control (Fig. 5.4, Fig. 5.5a). For the summer of 2011 (i.e., the start of the experiment through July 2011), limpet abundances were low in all treatments and summer emersion temperatures were high. From September to late March (winter), plate temperatures were cool and limpet abundance was high. From late April until August (summer 2012), plate temperatures were high during daytime low tides and limpet abundances were variable but mostly low (except for cool control plates). To explore patterns of community structure in each of these three phases of the experiment, I present data that corresponds to the end of each of these three seasons. At the end of summer 2011 (late July, Fig. 5.10a), warming modified the effect of herbivory on community structure (temperature × limpet: F1,28=3.61, p=0.009; Table 5.6) though the replicates within each treatment were similarly variable (PERMDISP: temperature × limpet:

F3,28=1.18, p=0.616; Table 5.6). Pairwise contrasts indicated that each of the four treatments were significantly different in location in MDS space (Table 5.6). Replicate communities within each of the cool treatments were composed of species that contributed consistently to similarity as indicated by high ratios of average similarity / SD (Table 5.9a, Table E.1a). By contrast, warm communities had less consistent contributions by constituent species. Diatoms were typical of all treatments, however they contributed less (45-54%) to the total similarity of cool communities, and more (78-87%) to warm communities. Cool communities were also characterized by higher B. glandula abundance, which contributed 25-37% to similarity. Green algae contributed to the difference in community structure between limpet treatments; Ulva sp. contributed 29% to community similarity in cool exclusions, while green filamentous algae contributed 12% to similarity in warm exclusions (Table 5.9a). The high contribution of green filaments was due largely to a single replicate plate with 100% cover of green filaments and 0% cover of diatoms. The remaining replicates had 100% diatom cover with 0-19% of greens compared to warm controls plates, which all had 0% greens.

120

Figure 5.10: nMDS plots highlighting the effect of warming and herbivory on community structure over the 16-month-long experiment on representative dates (see Results); (a) the end of the first summer, July 27, 2011; (b) the end of winter, April 9, 2012; and (c) the end of the second summer (and end of the experiment), August 28, 2012. Taxa percent cover was standardized to the total for each species prior to analysis. Warm treatments are represented by dark red symbols, and cool treatments by blue symbols. Limpet-controls are represented by solid circles, and limpet-exclusions by open squares. See Tables 5.5-5.7 for complete PERMANOVA results and Table 5.9 for SIMPER analysis. n = 8 for each treatment.

121 Source df SS Pseudo-F t P No. permutations Summer 2011 PERMANOVA temperature (t) 1 7,006 12.08 < 0.001 9,956 limpet (l) 1 2,787 4.81 < 0.001 9,955 t x l 1 2,091 3.61 0.009 9,960 residual 28 16,237 total 31 28,121 Pairwise comparisons warm con vs. cool con 1.67 0.041 3,977 warm excl vs. cool excl 3.52 < 0.001 4,009 warm con vs. warm excl 2.12 0.002 1,624 cool con vs. cool excl 1.81 0.030 5,004

PERMDISP t x l 3,28 1.18 0.616 9,999

Table 5.6: Effect of treatments on differences (PERMANOVA) and variability (PERMDISP) in community assemblage for July 27, 2011 (summer 2011). Pairwise comparisons are given when significant differences were detected.

Source df SS Pseudo-F t P No. permutations Winter 2012 PERMANOVA temperature (t) 1 2,282 1.28 0.280 9,943 limpet (l) 1 10,839 6.10 < 0.001 9,953 t x l 1 2,414 1.36 0.242 9,959 residual 28 49,765 total 31 65,299 Pairwise comparisons warm con vs. warm excl 1.92 0.005 5,066 cool con vs. cool excl 1.94 0.002 5,092

PERMDISP limpet 2,32 10.80 0.003 9,999

Table 5.7: Effect of treatments on differences (PERMANOVA) and variability (PERMDISP) in community assemblage for April 9, 2012 (winter 2012). Pairwise comparisons are given when significant differences were detected.

By the end of winter (early April), communities in temperature treatments converged, and only the main effect of the limpet treatment was significant (limpet: p<0.001, Fig. 5.6b, Table 5.7). In addition, limpet-control and exclusion treatments differed significantly in their dispersion

(PERMDISP: limpet: F2,32=10.80, p=0.003). The limpet-control treatments were more variable (Table 5.9b, Table E.1b); composed of several species that contributed inconsistently (Sim/SD 0.23-1.06) and relatively evenly (11-28%) to the low degree of similarity (34). The assemblage in limpet-exclusions was less taxonomically variable; the community was dominated by Ulva sp, which contributed to 70% of the total similarity (53) of exclusion communities. Though present

122 in both treatments, the asymmetry in Ulva sp. abundance was the leading contributor (18%) to dissimilarity between control and exclusion communities (avg. dissimilarity = 66; Table E.1b).

Source df SS Pseudo-F t P No. permutations Summer 2012 (a) PERMANOVA temperature (t) 1 10,396 13.56 < 0.001 9,957 limpet (l) 1 6,737 8.79 < 0.001 9,945 t x l 1 2,109 2.75 0.038 9,969 residual 28 21,465 total 31 40,706 Pairwise comparisons warm con vs. cool con 1.38 0.114 5,091 warm excl vs. cool excl 4.53 < 0.001 5,056 warm con vs. warm excl 3.31 < 0.001 5,033 cool con vs. cool excl 1.78 0.005 5,105

Treatment Bray-Curtis similarity (b) Average similarity among treatments cool excl 71.996 cool con 45.521 warm excl 66.588 warm con 72.661

Average dissimilarity between treatments warm con vs. cool con 54.772 warm excl vs. cool excl 44.519 warm con vs. warm excl 54.721 cool con vs. cool excl 51.897

Source df SS Pseudo-F t P No. permutations (c) PERMDISP t x l 3,28 6.50 0.004 9,999 Pairwise comparisons warm con vs. cool con 3.67 0.004 9,999 warm excl vs. cool excl 0.39 0.740 9,999 warm con vs. warm excl 0.51 0.648 9,999 cool con vs. cool excl 3.81 0.003 9,999 warm con vs. cool excl 0.20 0.855 9,999 warm excl vs. cool con 2.78 0.020 9,999

Treatment Mean ± SE (d) Average dispersion cool excl 18.374 ± 2.297 cool con 36.049 ± 4.030 warm excl 20.184 ± 4.048 warm con 17.608 ± 3.006

Table 5.8: Effect of treatments on differences in community structure for August 28, 2012 (end of experiment, summer 2012). Significant differences between communities are shown using (a) PERMANOVA with pairwise comparisons and the magnitude of difference can be extrapolated from the (b) average similarity and dissimilarity between treatments. Significant variability between replicate plates among treatments was determined using (c) PERMDISP, with pairwise comparisons. The magnitude of differences can be extrapolated from the (d) mean dispersion of each treatment.

123 Date / Treatment Species Av. Abund Av. Sim Sim / SD Contr. (%) Cum. (%) (a) Summer 2011 (July 27, 2011) Cool, con Diatom film 3.68 41.20 5.47 54.77 54.77 75.23 B. glandula 2.47 28.34 4.33 37.67 92.45

Cool, excl. Diatom film 3.90 36.71 4.49 45.34 45.34 80.97 Ulva sp. 2.50 23.65 3.23 29.21 74.55 B. glandula 2.31 20.61 2.99 25.45 100.00

Warm, con Diatom film 4.04 50.21 3.07 78.40 78.40 64.04 B. glandula 1.25 9.20 0.81 14.36 92.76

Warm, excl. Diatom film 4.04 59.67 1.63 87.68 87.68 68.05 Gr. fil. algae 1.49 8.22 0.65 12.08 99.76

(b) Winter (April 9, 2012) Control B. glandula 1.66 9.65 0.83 28.12 28.12 34.30 Ulva sp. 2.11 8.53 0.83 24.88 53.00 C. dalli 0.94 4.76 1.06 13.89 66.89 C. peregrina 1.18 3.93 0.69 11.47 78.35 Diatom film 1.41 3.78 0.42 11.03 89.38 Gr. fil. algae 0.93 1.48 0.23 4.30 93.69

Exclusion Ulva sp. 4.21 37.28 3.13 70.25 70.25 53.06 Navicula 1.54 5.33 0.51 10.04 80.29 Diatom film 1.36 3.14 0.37 5.92 86.21 P. fascia 0.75 2.86 0.45 5.40 91.60

(c) Summer 2012 (August 28, 2012) Cool, con B. glandula 2.76 25.17 1.55 55.28 55.28 45.52 Ulva sp. 2.29 11.04 0.71 24.26 79.54 C. dalli 1.45 6.62 0.66 14.54 94.08

Cool, excl. Ulva sp. 4.34 42.48 7.40 59.01 59.01 72.00 B. glandula 2.96 25.82 6.47 35.86 94.87

Warm, con B. glandula 2.38 20.97 2.25 28.85 28.85 72.66 Diatom film 2.61 18.65 1.59 25.67 54.52 Ulva sp. 2.11 17.96 3.67 24.72 79.25 C. dalli 1.97 14.87 2.23 20.47 99.71

Warm, excl. Diatom film 3.94 37.89 4.29 55.25 55.25 68.59 Ulva sp. 3.34 25.38 1.52 37.00 92.25

Table 5.9: Percentage contributions of individual species to observed similarity for each treatment, estimated using SIMPER analyses (when significant differences were detected using PERMANOVA, see Tables 5.6-5.8). The cumulative 80% of contributors to dissimilarities are shown. Sim / SD: the average contribution divided by the standard deviation of those contributions across all pairs of samples making up the average. Cont: contribution of each species to differences between treatments; Cum: running total of the contribution to the observed dissimilarity. Numbers beneath treatment names are the total similarity for that treatment.

124 By the end of summer 2012 (late August), when thermal stress was high and limpet abundance was variable, warming again modified the effect of herbivory on community structure (temperature × limpet: p=0.038, Fig. 5.10c, Table 5.8). The communities in each treatment also differed in dispersion (PERMDISP: temperature × limpet: F3,28=6.50, p=0.004; Table 5.8). Pairwise contrasts revealed that these effects were largely driven by the difference between cool controls and the other three treatments (Table 5.8). Replicate cool limpet-control plates were highly variable in community structure (similarity = 45) compared to replicates within other treatments (similarities = 69-72). Balanus glandula contributed to 55% of the similarity in cool controls, but replicates were variable with respect to Ulva sp., and C. dalli abundance (Sim/SD = 0.66-0.71). Warm controls, warm exclusions, and cool exclusions were equally variable. The community structure of warm controls was statistically indistinguishable from cool controls (Table 5.8), however replicate warm control plates were more similar to each other (similarity = 73), and B. glandula, diatom film, Ulva sp., and C. dalli contributed consistently and relatively equally to community structure (20-29% each; Table 5.9c, Table E.1c). Replicate communities in the warm limpet-exclusion treatment were highly similar (similarity = 69) and were dominated by diatoms and Ulva sp., which consistently contributed (Sim/SD = 1.5-4.3) to 93% of the similarity between replicates. Cool limpet-exclusion communities consisted almost exclusively of Ulva sp. and B. glandula, which were highly consistent contributors (Sim/SD = 6.5-7.4) to 95% of the total similarity (similarity = 72; Table 5.9c).

5.4.6 Community trajectory

Warming modified the effect of limpet herbivory on the trajectory of community development over 16 months (Fig. 5.11, Table 5.10), and pairwise comparisons revealed significant differences between each of the average treatment trajectories (Table 5.10). Communities in warm treatments did not change as much as communities in cool treatments initially (shorter distance from time 0 to time 1). Over the first summer (points 0-4, Fig. 5.11), warm controls and exclusions followed a similar path, and the trajectories in cool controls and exclusions were also similar to one another. Later that summer, the average community in the warm exclusions deviated from the other treatments due to higher abundances of green filaments (data not shown). By the end of summer, communities were significantly different from each other (point 4, Fig. 5.11; Fig. 5.10a). From the end of summer to the end of winter (points 4 to 7, Fig. 5.11), the

125 community changed more in warm exclusions, resulting in very similar communities in both exclusion treatments (point 7, Fig. 5.11). Through early summer 2012 (points 8 to 10, Fig. 5.11), community change on warm and cool plates within each of the limpet treatments was similar. However, in early July (Fig. 5.11, point 10), the communities in warm and cool exclusions diverged sharply; the community in cool exclusions remained Ulva- and B. glandula-dominated while the community in warm exclusions reverted to a ‘winter’-like community, dominated by diatoms. In limpet-controls, warm and cool plates continued to follow similar trajectories over summer, both nearing their respective ‘winter’-like states by August (points 7 and 12 are close to each other in each of the warm and cool controls, Fig. 5.11). To determine whether significant differences between trajectories (above) were due to difference in average trajectory (location) or to the variability between replicate plates (dispersion), a PERMDISP analysis was conducted on the second stage results and visualized in nMDS space (Fig. 5.12). Each point represents the entire trajectory (over 16 months) of a single replicate plate, and points closer together indicate more similar trajectories. Warming modified the effect of herbivory on the variability in community trajectory (Fig. 5.12, Table 5.10). Pairwise comparisons revealed that replicates within each of the cool limpet-controls and warm limpet-controls were highly variable, but the two treatments were similarly variable (solid blue circles as dispersed as solid red circles; pairwise comparison: p=0.524). By contrast, replicate plate trajectories within each of the cool exclusions and warm exclusions were less variable than controls, but the two treatments were similar in variability to each other (open blue squares as clustered as open red squares; pairwise comparison: p=0.766). Thus, the overall difference in recovery trajectories between warm and cool controls and between warm and cool exclusions was due to a difference in the average trajectory (e.g., location), and not to variability (e.g., dispersion). The variability (dispersion of points within each treatment) between the remaining four combinations of treatments was significantly different (e.g., cool exclusion replicates had less variability than cool control replicates; Table 5.10). Thus, differences between warm exclusions and controls and between cool exclusions and controls may have been due to either the average trajectories or the variability in trajectories or both.

126

Figure 5.11: (a) nMDS of community development over time for each treatment. The distance between data points in the ordination represents the difference in average community structure between time points. The black symbol represents the start of the experiment (blank plates on May 3, 2011). The full PERMANOVA analysis with pairwise comparisons can be found in Table 5.10. Small numbers within symbols correspond to each sampling date (b), and representative dates in Fig. 5.10 can be identified from the table. Note that sampling dates are not equally spaced through time.

127 Source df SS Pseudo-F t P No. permutations (a) PERMANOVA temp (t) 1 1.40 7.08 < 0.001 9,945 limpet (l) 1 0.64 3.21 0.002 9,948 t x l 1 0.46 2.34 0.023 9,944 residual 28 5.54 total 31 8.04 Pairwise comparisons warm con vs. cool con 1.98 0.001 5,020 warm excl vs. cool excl 2.55 < 0.001 5,043 warm con vs. warm excl 1.61 0.015 5,076 cool con vs. cool excl 1.71 0.018 5,082

Treatment correlation coefficient (b) Average similarity among treatments cool excl 0.551 cool con 0.249 warm excl 0.520 warm con 0.307

Average similarity between treatments warm con vs. cool con 0.148 warm excl vs. cool excl 0.378 warm con vs. warm excl 0.350 cool con vs. cool excl 0.323

Source df SS Pseudo-F t P No. permutations (c) PERMDISP t x l 3,28 6.43 0.006 9,999 Pairwise comparisons warm con vs. cool con 0.74 0.524 9,999 warm excl vs. cool excl 0.35 0.766 9,999 warm con vs. warm excl 2.73 0.033 9,999 cool con vs. cool excl 3.35 0.012 9,999 warm con vs. cool excl 2.77 0.037 9,999 warm excl vs. cool con 3.38 0.013 9,999

Table 5.10: Effect of treatments on differences in successional trajectories among treatments using (a) PERMANOVA with pairwise comparisons. The magnitude of the difference can be extrapolated using the (b) average similarity and dissimilarity between treatments (estimated with correlations coefficients). Significant variability between replicate plates among treatments was determined using (c) PERMDISP, with pairwise comparisons. Data were log(x+1) transformed with a Bray-Curtis resemblance matrix, 9999 permutations. All terms were fixed, using Type III SS.

128

Figure 5.12: nMDS plot of successional patterns through time for each replicate plate over 16 months. Each point represents the trajectory for each plate, specifically, the correlation structure between time points for a single plate that are independent of the identity of taxa. Differences between points show whether the trajectory over time was similar or variable among replicate plates (n = 8) for each treatment. Complete analyses can be found in Table 5.10.

5.5 Discussion

Climate change has long been known to affect the distribution of species and the structure of ecological assemblages. In the late Pleistocene, climate change resulted in the compositional turnover of species in ecological communities (Graham and Grimm 1990, Roy et al. 1995, Graham et al. 1996). Likewise, anthropogenic climate change is expected to cause changes in community structure and composition because species interactions will shift due to direct effects on performance, phenology, and vertical and latitudinal distribution (Hughes 2000). Temperature directly determines physiological processes for all species, and thus indirectly affects population growth, species interactions, and community dynamics. The more quantitatively we can parameterize these linkages, the better we can understand and predict warming induced changes in natural systems. As mean temperatures and variability increase, species will be more likely to experience temperatures higher than their optimum, with important implications for performance (Vasseur et al. 2014). Thus understanding how thermal stress affects species interactions and communities will be key.

129 5.5.1 Efficacy of warming and herbivore removals

Black and white settlement plates and copper exclusions were effective in manipulating temperature and herbivore density, although the strength of treatment effects varied seasonally. Initially, warm summer temperatures created distinct thermal regimes on black and white plates, while limpet density was low. In winter, daytime high tides eliminated thermal differences, and limpets became abundant, creating discrete limpet treatments. In summer 2012, warming and moderate limpet densities created four distinct treatments. The seasonality of these treatments was apparent in the responses of many individual taxa as well as overall patterns of succession and community structure. Projections estimate that the global mean temperature will increase by 1.0-3.7°C over the next century, a change that will be accompanied by increased variability (IPCC 2013). Both of these metrics influence how species will respond to warming (Vasseur et al. 2014), so it is important that climate change experiments incorporate both. In the present study, I used black and white settlement plates to change the thermal regime by increasing the mean, maximum, and variance in substratum temperature. Black plates were 1.6°C hotter than white plates on average, which is consistent with previous experiments using similar methods (Chapter 3, Lathlean and Minchinton 2012). This design took advantage of solar radiation to warm the substratum, thus only increased temperature during daytime and when plates were exposed at low tide. Air temperatures are expected to increase faster than water temperatures (Sutton et al. 2007), and intertidal organisms experience increased thermal stress during low tide (Hofmann and Somero 1995). Further, many prostrate intertidal organisms have body temperatures that are highly correlated to substratum temperatures (Denny and Harley 2006). Thus, this in situ heating method effectively increased stressful temperatures for intertidal species (see Chapter 3 for discussion of caveats of this warming method). Copper sheeting and warming reduced the density of the dominant mid intertidal herbivore on plates, although the effect varied over time. Limpets did not appear in plate surveys for the first eight weeks of the experiment, although they may have been making feeding forays on the plates at high tide or at night during this time, when conditions were less stressful. Limpets were first recorded in daylight, low-tide surveys late in the first summer, became abundant in controls in winter, and were variable (over time and between treatments) the following summer. Lottia pelta and L. scutum were the most abundant limpets found on plates,

130 and both have thermal tolerances near the maximum rock temperatures I measured. Hsp70 expression begins at a substratum temperature of 28°C in both species (Dong et al. 2008), and thermal lethal limits for L. pelta and L. scutum are 39°C and 40°C, respectively (Wolcott 1973). In the current experiment, the average absolute maximum temperature was 34.5°C on white plates and 40.5°C on black plates, thus black plates were likely thermally stressful and avoided more often by limpets, especially in summer. On most dates, warming and exclusions reduced limpet density in a multiplicative fashion, however in the three surveys where limpets were most abundant compared to other dates, the combined effect of warming and exclusion was synergistic. Since algae (diatoms and Ulva) were abundant in warm exclusion treatments at these times, it is unlikely that food was limiting for limpets. One possible explanation is that barnacles were also nearly absent. Adult barnacles create a more topographically complex habitat, which reduces desiccation, and limpets are often observed ‘nestled’ among them during low tide (Berlow and Navarrete 1997). Without barnacles, warm limpet-exclusions may have been even more inhospitable than would be expected.

5.5.2 Independent effects of herbivores and warming on taxa

Limpet presence and warming independently affected population responses via a combination of direct effects on specific taxa and indirect effects mediated by trophic and competitive interactions. Limpets negatively affected Ulva and diatom cover, presumably through direct consumption of microscopic stages (Castenholz 1961, Nicotri 1977). Limpets had negative and presumably indirect effects on other grazers (e.g., amphipods and Littorina sp.) and mobile predators (e.g., Emplectonema gracile), although the effects were small. Many mobile invertebrates require the shelter provided by later succession upright species such as mussels (Hewatt 1935), and when mussels are rare, foliose algae could serve a similar function. Mobile invertebrates were very rare in all treatments until summer 2012. During that time, limpets reduced total algal cover by ~75% (in cool treatments), eliminating potential habitat for mobile species. In addition, limpets and littorine snails have very similar diets (Nicotri 1977) and limpets can outcompete adult littorines for resources (Harley 2002). Limpets indirectly positively affected barnacles (B. glandula and C. dalli), particularly early in both summers when barnacle larvae were settling. At a coarse scale, settling barnacle

131 larvae are attracted to diatoms and avoid green filamentous algae, because the former are associated with moist habitats while the latter are associated with higher tidal heights (Le Tourneux and Bourget 1988). At a finer scale, larvae select completely clean sites on which to permanently adhere (Le Tourneux and Bourget 1988). Initially, B. glandula settled where diatoms were present (cool treatments), but cover may have become too thick in cool exclusions (73-80% cover) to allow for additional settlement. In the second summer, limpet presence was also positively associated with B. glandula settlement, although this was likely mediated through the reduction in Ulva cover (~85% less in cool controls), providing bare space for settlement. The same mechanism was likely responsible for the positive effect of limpets on encrusting algae in winter and spring. Warming affected taxa via a combination of direct thermal effects and indirect effects mediated by competitive interactions between major space occupiers. Warming directly negatively affected B. glandula density but positively affected C. dalli. In similar experiments, I found that warming reduced B. glandula post-settlement survivorship by 87%, leading to a smaller population after one year (Chapter 4, Table C.1). At 33°C, B. glandula exhibit a strong intracellular stress response (Berger and Emlet 2007) and 46°C is the lethal thermal limit for adults (Liao & Harley, unpublished). The average maximum temperatures recorded on my plates were well within this hazardous range (34°C on white and 40°C on black), so it is likely that warming from black plates was stressful for B. glandula. C. dalli has a higher thermal lethal limit than B. glandula (Liao & Harley, unpublished). Though C. dalli settle throughout the intertidal, they typically only persist in the highest zone on Salt Spring Island because mortality rates lower on the shore approach 100% (Chapter 4), likely due to biological interactions. Although the effect was small, C. dalli was indirectly benefited by the warm treatment because it was released from competition with B. glandula. By the second summer, all plates were too densely settled for C. dalli to recruit. Ulva, diatoms, and green filaments were also affected by warming, likely through a combination of direct effects mediated by thermal tolerance and indirect effects mediated by competitive interactions. The temperature treatment affected diatoms in different ways at different times of the year, due in part to complex interactions with other species. In early succession (summer 2011), the timing and abundance of diatoms and green filaments was dependent on plate temperature. Diatom recruitment was delayed and green filamentous algae

132 bloomed on warm plates. Whether warming directly inhibited diatoms, allowing for green filament settlement or warming promoted green filament settlement, delaying diatoms, is difficult to determine without additional experiments. In previous, similar experiments, I did not observe an effect of the temperature treatment on green filament abundance (in the absence of heavy diatom cover; Chapter 3). Thus, filamentous greens may have opportunistically occupied the free space provided by the delay in heavy levels of diatom establishment in this experiment. However, this suggests that diatoms were negatively affected by warming, which is inconsistent with the spike in diatom cover on warm plates in July. Since neither seemed to be directly affected by the warming treatment, there may have been some indirect pathway leading to the apparent warming induced switch in abundance in the first summer. Ulva cover was reduced in the warm treatment (in exclusions) in early winter and the second summer, although the mechanism was less clear. Ulva began recruiting on cool (exclusion) plates earlier than on warm (exclusion) plates, despite very high diatom cover in both treatments. This may have been due to (1) cooler temperatures on cool plates and / or (2) facilitation by B. glandula, which was absent on warm plates. In both treatments, diatom cover decreased as Ulva cover increased in late winter. In the second summer, warming reduced Ulva cover by ~30%, allowing space for diatoms to colonize.

5.5.3 Warming modifies species interactions

Warming can weaken, strengthen, or reverse the sign of species interactions by changing the relative performance and interdependencies of interacting species. On Salt Spring Island, warming modified the effect of limpets on diatoms and B. glandula and these effects changed over the 16 months of the experiment. This may have been due to seasonally changing abiotic conditions, the ontogeny of interacting species (Kordas and Dudgeon 2011), or interactions with other community members. Although I did not explicitly quantify the interactions between all pairs of species, the relative patterns of species abundance provide possible explanations for the temporal changes in thermal dependent interactions. The response of diatoms to warming and limpets varied over time, likely due to release from competitive interactions with other taxa. Diatoms are directly consumed, purposefully and incidentally, by many intertidal herbivores (Steneck and Watling 1982, Riera et al. 2004). They are also opportunistic colonizers; diatoms reproduce continually and grow quickly, so they can

133 colonize bare space whenever it becomes available. Thus diatom populations may be highly variable and highly dependent on the demography, vital rates, and behavior of several species simultaneously. For example, benthic diatoms act as a settlement cue for barnacles, but in high density, can prevent settlement (Le Tourneux and Bourget 1988). However, if barnacles pre- empt space, diatoms can still settle thickly on large barnacle tests with little negative effect to barnacles (RLK personal observation). Thus interactions between diatoms and other species (such as barnacles) likely depend on the density of diatoms and the relative timing of settlement of interacting species, which has been observed in other algal-invertebrate pairs (Kordas and Dudgeon 2011). In this study, the effect of limpets on diatoms was weakened or reversed by warming at various times. For example, in September, limpets positively affected diatom cover but warming reversed this effect. Since plates did not differ in temperature at this time, the effect was likely indirectly mediated through barnacles. Balanus glandula were large and abundant in cool controls (personal observation), which may have reduced grazing ability by limpets (Farrell 1991), keeping diatom cover at 30%. In warm controls, B. glandula were less abundant and diatom cover was 2%. With fewer barnacles, grazing limpets may have had easier access to diatoms. In late summer 2012, a different mechanism was likely responsible for the reversed effect of limpets in warm treatments. Diatom cover was similarly low (under 3%) in both cool treatments despite eight times more limpets in controls and four times higher Ulva cover in exclusions. At the same time, all species of (non-limpet) grazers were more abundant in cool exclusions, likely due to a combination of reduced competition with limpets and increased habitat (Ulva). Littorine snails, amphipods, and Idotea consume diatoms and other microalgae (Steneck and Watling 1982, Ruesink 2000, Mancinelli 2012). Thus, in cool controls, limpets likely consumed diatoms (and Ulva) directly. In cool exclusions, other grazers benefited from the absence of limpets, and directly consumed diatoms. In warm controls, limpet directly consumed diatoms (and probably Ulva). However, other grazers did not benefit from the absence of limpets in warm exclusions. This may have been because higher temperatures reduced habitat complexity (Ulva occupied half as much space in warm exclusions compared cool exclusions) or were physiologically stressful for grazers. Given that the thermal optima for many marine benthic diatoms exceeds 25°C (Admiraal 1977, Admiraal and Peletier 1980), while that of Ulva is closer to 16°C (Steffensen 1976), it is not surprising that diatoms would fare better than Ulva in warm exclusions.

134 Thermal stress strengthened the indirect positive effect of limpets on B. glandula. Limpets positively affected B. glandula density in cool and warm treatments, likely by consuming algae (diatoms and Ulva), which reduced competition (e.g., provided free space for settlement or less interference with their feeding apparatus) for settling barnacles (Le Tourneux and Bourget 1988). Overall, algal cover was higher in each warm treatment (largely due to greens and diatoms) relative to cool treatments. In warm exclusions, algal cover was high since grazing pressure was nearly absent. Thus, competition with algae and increased physiological thermal stress drastically reduced B. glandula density in warm exclusions. Balanus glandula fared better in warm controls than warm exclusions, due to the release of algal competition. In addition, the greater positive effect of limpets in warm treatments compared to cool treatments may have been dependent on barnacle density. In cool limpet-controls, B. glandula density was extremely high in May-June of both summers (400-500 individuals per plate), and reached 100% cover in mid summer 2012, thus the population likely experienced density dependent mortality under benign conditions (Bertness 1989). This was reduced in warm controls because barnacles were half as abundant, only reached 60% cover in summer 2012 (when the warming × limpet effect was strongest; Fig. 5.8b), and had reduced growth rates (Chapter 4). All species exist within a community of interacting pairs. Each species, and by extension, each interaction, is uniquely affected by temperature. Limpets directly consumed diatoms and indirectly benefited barnacles via reduced competition with algae. However, when the community was warmed, these pairwise interactions were strengthened, weakened, or reversed because non-target species were also differentially affected by limpets and warming (see Independent effects of herbivores and warming on taxa above). Manipulating temperature and interactions in a community context provides the opportunity to uncover indirect responses.

5.5.4 Community structure

Warming modified the effect of limpet presence on community structure, though the influence of each variable varied in relation to when each main treatment effect was strongest. Limpets had an important structuring influence on the assemblage throughout the experiment, despite temporally variable densities. The interaction between warming and limpet presence became important in summers when thermal stress was high. Since diatoms and B. glandula were large contributors to community structure, temperature dependent interactions (see Species

135 Interactions above) may have caused community level changes. Diatoms and B. glandula may therefore have acted as leverage points, causing communities to diverge among all treatments. In summer 2011, limpets were most effective at removing foliose algae, warming was most effective at reducing barnacle cover, and diatoms reacted opportunistically to both treatments. Although limpets were rarely observed in surveys in the first summer, it is likely that they were still grazing plates. Limpet access to plates significantly changed the community structure by eliminating green algae (Ulva in cool and green filaments in warm treatments). This reduced competition for juvenile B. glandula, which was consistent with my hypotheses. Warming depressed B. glandula and foliose algal (Ulva) cover, allowing more free space for diatoms. Although warming was detrimental for B. glandula, reduced algal competition via limpet consumption allowed B. glandula to persist in a diatom-dominated community in warm controls. These results confirm my predictions. Based on previous experiments in this region using the same method, I hypothesized that warming alone would shift community structure away from a barnacle-dominated state (Chapter 3). This pattern arose after four months (May-August) of summer warming in 2009 (Chapter 3), which is comparable to the July 2011 results (above). Consistent with previous results, B. glandula was a leading contributor to the difference between cool and warm controls (SIMPER pairwise analysis of cool and warm controls not shown); warming reduced B. glandula abundance, which freed space for other species. In the previous experiment, other species did not occupy this space, while in the current experiment, diatoms opportunistically settled onto bare substratum. The reason for this difference is unknown, but could involve differences between years, experimental sites, or subtle differences in shore level between the two experiments. In winter, the temperature differences between plates disappeared, although the difference in grazing pressure between limpet-controls and exclusions widened. This caused communities in cool and warm treatments to converge and communities in controls and exclusions to diverge. Consistent with my hypothesis, limpet grazing reduced diatom cover, which facilitated barnacles. In addition, limpets increased variability in community structure, in effect, increasing patchiness. Some replicate communities were more barnacle-dominated while others were more Ulva-dominated (i.e., each species in the control treatment had a low degree of

136 similarity). In the absence of limpets, communities were more similar, and consistently dominated by Ulva. The combined effect of warming and limpets on community structure is best determined from the last time point, months after settlement of many organisms and after months of influence from both treatments. Through summer (2012), temperatures diverged between black and white plates and limpet density was variable through time and between treatments. By August, the significant interaction between the warming treatment and the limpet treatment resulted from greater dissimilarity between cool and warm limpet-controls compared to warm and cool limpet-exclusions, which was largely due to differences in community variability. In cool treatments, limpets caused community structure to be patchy, with replicate plates dominated by B. glandula or Ulva, while others had a more even balance of both. By contrast, cool exclusions were very similar and Ulva-dominated. The range of assemblage types in controls was potentially influenced by periodic disturbance by limpets that resulted in plates being at several different stages of succession. In warm treatments, the addition of limpets did not change variability but did change the average structure, compared to exclusions. Limpets led to the development of a B. glandula-diatom community compared to a diatom-Ulva community in warm exclusions. Interestingly, cool and warm limpet-controls differed significantly in community variability but not in average community structure. Yet the communities in each were still highly dissimilar. This implies that limpets promoted a similar average community structure (barnacle-dominated) in both temperature treatments, however warmer temperatures decreased the variability of the final community type and resulted in more even species contributions to structure. This was likely due to the negative influence of thermal stress on B. glandula (which were less abundant in warm compared to cool controls) and indirect effects on diatoms (which were more abundant in warm controls compared cool controls due to more bare space, but less abundant than in warm exclusions due to increased grazing; discussed above). Overall, limpet presence shifted communities in the same direction - towards a higher contribution by B. glandula. Warming also shifted communities in the same direction - towards a higher contribution by diatoms. However, warming ultimately eliminated the variability- inducing effect of limpets, making final communities more similar.

137 5.5.5 Diversity

Warming and limpet presence modified diversity independently. When treatment effects were significant, they varied over 16 months, with the largest differences in late summer (2011) and early winter. During that time, warming significantly reduced diversity and evenness but species richness was similar. In July, diatoms dominated the community in warm treatments, contributing 83% to the similarity in community structure (SIMPER of warm vs. cool; data not shown). Cool treatments were more even, with species that contributed 17-50% to similarity. There was a trend towards reduced diversity and richness in warm exclusions (in 2011), however the effect was small (e.g., 2 versus 3 species) and not statistically significant. Warming modified the effect of limpets on communities predominantly via changes to the assemblage structure, rather than changes to overall diversity, richness, or evenness.

5.5.6 Community trajectory

The temporal dynamics of these intertidal communities in a warmer future depend on trophic interactions. The high variability in community trajectory characteristic of cool controls (compared to cool limpet-exclusions) demonstrates that herbivory promoted diverse patterns of succession. Limpet grazing generally reduced algal cover, allowing other species like B. glandula to become dominant on some replicates. The wide range of assemblage types (in winter and summer 2012) were due to periodic disturbances associated with limpets that resulted in plates being at several different stages of succession. For example, two replicates passed through a state nearly devoid of any organisms (‘bare substratum state’; data not shown), but because these disturbances occurred at different times (November 2011 and July 2012), the subsequent succession of each followed a different path. The November disturbance was followed by a bloom of Ulva then dense settlement of B. glandula the following spring. The July disturbance occurred after the peak of B. glandula settlement, thus barnacle density stayed low and little else recruited into the free space in the remaining six weeks of the experiment. It is not possible to definitively know what caused these disturbances (e.g., limpet grazing, sea star predation, log damage), but it may also be possible that limpets facilitated a community type that was more susceptible to disturbance.

138 Replicate warmed communities were no more variable in succession than comparable cool communities, however they did differ in the average pattern of succession. In general, warming caused communities to be dominated more often by benthic diatoms and green filamentous algae rather than habitat forming species like barnacles and Ulva. In warm exclusions, initial barnacle recruitment was variable between replicates, but by June-July 2011, nearly all barnacles on all replicates had died. Contemporaneously, green filamentous algae bloomed for 4-8 weeks, which was replaced by diatoms, and finally a mix of diatoms and Ulva in the second summer. Replicate plates of the cool exclusions treatment also followed a very consistent pattern of succession, however it differed from that on warm exclusions in terms of the frequency of species turnover. With limpets absent, warm plates experienced punctuated peaks by 3-4 species over 16 months, while communities on cool plates were dominated by a mix of diatoms and barnacles for the first half of the experiment then Ulva and barnacles for the second half of the experiment. Succession in warm and cool controls was similar in variability, but was statistically different in average trajectory. Patterns of succession were more dissimilar between warm and cool controls compared to warm and cool exclusions, leading to the significant temperature × limpet interaction. Communities in warm limpet-controls experienced both periodic disturbances associated with limpets and species turnover from warming. Similar to warm exclusions, replicate warm control plates began succession with varying levels of B. glandula recruitment. However, warm controls also had varying levels of barnacle mortality in June-July 2011, whereas exclusions had nearly complete mortality. Probably due to this variability in B. glandula abundance in summer 2011, successional paths diverged on replicates and were somewhat stochastic through winter depending on disturbance. For example, when barnacles were less abundant on warm controls, the immediate successional pattern mirrored that of warm exclusions (green filamentous algae followed by diatoms). However, when barnacles were initially more abundant on warm controls, the green filamentous bloom stage was skipped. In winter, limpets drove overall algal abundance below ~40% on five out of eight replicates, which allowed for high recruitment of B. glandula. On the remaining three replicates, algal abundance stayed high and barnacle abundance stayed low through the following summer. No replicate plates returned to a ‘bare substratum’ state as some cool control plates did. Instead, abundance declines of one species were always associated with a peak in abundance of another species, similar to warm

139 exclusions. However, the order of species’ peaks was not as predictable in warm controls as it was in warm exclusions, especially in winter and the following summer. Nevertheless, after 16 months, replicate warm control communities on these variable trajectories ended up relatively similar to each other in structure (compared to cool controls; Fig. 5.10c), with several species contributing consistently. Overall, limpets either directly created or increased the likelihood of stochastic disturbances that reset the successional trajectory, which resulted in increased variability among replicates over time. In general, warming altered the early succession of communities and promoted species turnover which may have been due either to higher mean temperatures or increased thermal variability. Warming altered the average community trajectory more when limpets were present due to the combined effect of disturbances and species turnover.

5.5.7 Summary and implications

Warming and limpets influenced many individual taxa independently, however warming modified the effect of herbivory on key species, which acted as leverage points in the assemblage. The influence of temperature on species interactions has recently gained popularity in the ecological literature. There is no apparent trend in the effect of temperature on interactions with respect to the type of interaction (e.g. herbivory), taxa, or habitat (O’Connor et al, unpublished data). Competition, facilitation, herbivory, and predation can weaken, strengthen, or be unaffected by warming (Burnaford 2004, Pincebourde et al. 2008, O'Connor 2009, Yamane and Gilman 2009, Lang et al. 2011, Moise and Henry 2012, Wang et al. 2012, Alsterberg et al. 2013, Lemoine et al. 2013). The relative changes in performance of interacting species determines whether interspecific relationships change in sign or magnitude as temperatures rise, which can have very different effects on community dynamics. Examining these interactions in terms of each species’ thermal performance curve has shed light on how changes in interactions are directly related to temperature (Fey and Cottingham 2012). However performance is also indirectly affected by temperature. Warming can influence interactions by affecting organismal behavior or habitat use (Barton 2010), the relative timing of demographic events (Rudolf and Singh 2013), and interactions with non-target species in the community (Barton and Schmitz 2009). Manipulating warming in situ, and in a community context allows the relative performance of interacting species to vary naturally with respect to these factors.

140 A small but growing number of studies have manipulated warming and herbivory simultaneously to determine their relative effects on community dynamics. In field manipulations, temperature and herbivory have had independent effects on total algal abundance, community structure, and diversity (Williams 1994, Thompson et al. 2004, Dethier et al. 2005, Morelissen and Harley 2007, Tamburello et al. 2013, Williams et al. 2013), but the interaction between temperature and herbivory was not significant in any of these studies. There are probably several possible reasons for this, however one may be that communities were not thermally stressed. These experiments were either in thermally benign locations and/or used shades to reduce temperature, rather than increasing it, so the strength of herbivory may not have responded to cooler temperatures (which is not to say that resources and herbivores did not respond, but rather that they did not respond at different rates). In thermal ‘hot spots’, warming due to climate change will push intertidal organisms (which already live near their thermal max) past their thermal tolerances. Thus manipulations that increase stress rather than decrease it may be more applicable to future thermal conditions. Warm treatments in this experiment were several degrees hotter than cool treatments and significantly more variable, which reflects near future thermal changes predicted from climate change. The significant interaction between warming and herbivory observed here (but not elsewhere) may have been a reaction to thermally stressful conditions. As this study demonstrated, species interactions are of fundamental importance to the response of communities to climate change. Warming and herbivores together increased abiotic and biotic pressure on dominant species, resulting in markedly different trajectories due to herbivore disturbance and high species turnover. Despite the stochastic nature of development, these communities ultimately lost the variability created by herbivore associated-disturbances, resulting in highly similar assemblages. These results suggest that future warming may cause a change in the functional role of dominant herbivores on community dynamics. Although increasing temperature will simultaneously affect many aspects of an ecological assemblage, disentangling the two is key to understanding the mechanisms underlying ongoing changes and developing more accurate predictive models of how key species may mitigate or exacerbate the effects of warming.

141 Chapter 6 Concluding Remarks

6.1 Synopsis

As the Earth’s climate changes so too do its ecosystems, due to shifts in abundance, biodiversity and interaction strengths among their constituent species. As temperatures rise, interactions such as competition, predation, and herbivory are changing due to shifts in per capita interaction strength and the relative abundance of interacting species. Changes in interspecific relationships, in turn, can drive important local-scale changes in community dynamics, biodiversity, and ecosystem functioning, and can potentially alter large-scale patterns of distribution and abundance. The combination of species-specific thermal responses, complex population dynamics, and context dependent interaction strengths can make it challenging to predict how populations and communities will respond to climate change (Fig. 6.1). Nonetheless, an improved understanding of how assemblages of multiple, interacting species will respond to climate change is imperative if we hope to effectively prepare for and adapt to its effects. The primary goal of this dissertation was to determine how warming affects community assemblages via direct (mediated by organismal physiology) vs. indirect effects (mediated by species interactions). In Chapter 2 I presented a review of the current literature focused on how individual thermal responses can determine species interactions. A secondary goal of this dissertation was to improve upon current field-based methods for manipulating temperature. I achieved this by creating heated settlement plates. In the remaining chapters, I used this experimental approach to examine direct and indirect responses to warming. In Chapter 3 I examined the functional group and community level patterns that resulted from one year of substratum warming. In Chapter 4 I took a more detailed look at how warming affected organismal vital rates and population size to better understand direct contributions to community level patterns observed in Chapter 3. In Chapter 5, I investigated whether the effect of species interactions on community dynamics depended on environmental temperature. The main conclusions of the research are outlined below. I will begin by addressing my secondary goal. I will discuss the importance of experimental warming manipulations and will include specific advantages and disadvantages associated with the methods used for these experiments. Finally I

142 will finish with the main ecological conclusions and recommendations for future research directions.

Figure 6.1: Conceptual diagram showing the pathways by which climate warming affects direct and indirect processes, ultimately determining community dynamics. Each color (orange, yellow, green, blue) represents a different species. Boxes with touching corners feed directly into each other: temperature affects metabolic rates, which determine rates of resource acquisition (e.g., carbon, nutrients, or prey). The ability of organisms to attain resources determines the interaction strength (IS) between individuals (per capita IS, solid arrows) and the growth and reproduction rates of the population. Relative population sizes and density dependent (population IS, solid arrows) interactions (which can feedback to modify population size, dashed arrows) determine the structure and dynamics of ecological communities. Modified from Chapter 2 (Kordas et al. 2011) and Harley (2013) to include community-level dynamics.

143 6.2 Improving realism for global warming experiments

6.2.1 Classic warming experiments

Warming is arguably the most important facet of global change because temperature affects nearly all biological processes in all habitats. Early investigations concerning the effects of warming, often using a single species and conducted in the laboratory, found variable responses. However, these studies lack the realism found in natural systems, where species interactions and secondary adaptations (e.g. behavioral responses) can supersede the importance of individual tolerances, emphasizing the need for field-based community studies. Because of logistical limitations, field-based community-level warming manipulations are still relatively rare, especially in marine environments (Wernberg et al. 2012a). However, they are desperately needed; as Wernberg (2012a) puts it, “indeed the challenge for experimental ecologists is to provide real and realistic data that will expand the collective inference space of our mechanistic understanding of how the physical forcing of climate change translates into ecological changes” Much has been learned from two main types of marine field-based thermal stress studies, those that substitute space-for-time (Leonard 2000, Wernberg et al. 2010) and those that use experimental shades in the intertidal (Underwood 1980, Wethey 1984a). Space-for-time experiments use locations that differ in an environmental variable of interest as a substitute for different time periods (i.e., current and future warming). Unfortunately it is nearly impossible to find identical locations that only differ in the variable of interest. Traditionally, in situ experiments manipulating thermal stress, have employed intertidal shades (Underwood 1980, Wethey 1984a, Morelissen and Harley 2007). The results from shading experiments are mixed; macroalgae benefited (Williams 1994, Dethier et al. 2005, Sampath-Wiley et al. 2008), were harmed (Kavanaugh et al. 2009), and were unaffected (Bertness and Leonard 1997) by shading. Under naturally hot conditions, barnacles faired better under shades (Bertness 1989, Bertness et al. 1999a) unless temperature sensitive predators were present (Harley and Lopez 2003). In many shading experiments, invertebrate grazers responded positively (Tomanek and Sanford 2003, Burnaford 2004), although in some experiments grazers did not respond (Bertness et al. 1999b, Morelissen and Harley 2007). Thus, the response of marine species and communities to warming has been inconsistent and may be due in part to the method itself.

144 Shades have the unintended consequence of reducing light irradiance, changing water motion, and providing shelter for mobile invertebrates (Underwood 1980, Miller and Gaylord 2007). These artefacts may change species interactions and community structure (primary productivity, grazing pressure), thus results from shading experiments must be interpreted with caution. They are particularly problematic when used to infer non-linear biological responses to climate warming, because shades do not mimic warming but cool one treatment instead (Fig. 6.2). For example, 5°C of cooling may not elicit the same change in a biological response as 5°C of warming, especially for organisms or ecosystems near a threshold or tipping point, where small increases in temperature can be lethal. Further, this comparison may be misleading because organisms are adapted to ambient conditions

Figure 6.2: Population growth rate of the marine diatom Phaeodactylum tricornutum (Kudo et al., 2000). When using treatments that cool (e.g., shades) rather than heat to infer potential effects of climate warming, the implicit assumption is that performance increases linearly with temperature, so a decrease in performance when cooled should correspond to an increase in performance when warmed (a). However, if ambient habitat temperature is close to the thermal optimum for a species, performance will decrease sharply as temperatures rise, thus extrapolations using cooled treatments will misrepresent performance in a warmer future (b).

6.2.2 Advantages and disadvantages of passive solar heated settlement plates

This solar heated plate design allowed me to gain insights that would not have been obtainable using other approaches. However, like other heated plate designs (Smale et al. 2011, Lathlean et al. 2013) and indeed all in situ warming methods, this technique had advantages and disadvantages. The most distinctive advantage was the ability to mimic future temperature changes. Heated plates are a significant improvement on other field-based temperature

145 manipulations (e.g., shades) because conductance-mediated methods warm the substratum rather than cooling it and don’t affect the light regime (Lathlean and Minchinton 2012, Smale and Wernberg 2012). A drawback to many lab, mesocosm, and even some field manipulations of temperature is the inability to simulate a realistic temperature regime. Natural weather-induced variability can be homogenized by treatments that maintain constant temperatures, and the difference between treatments may not reflect climate change scenarios. Both of these limitations were overcome in this experiment. The plates were solar heated, so the surface temperature tracked the local climate and weather, providing a realistic in situ change in temperature. Although experimentally warming the substrate is not the same thing as climatic warming of the air, the effects on body temperature are extremely similar for the species considered in this thesis. For example, although substratum and limpet body temperatures tend to exceed air temperature by several degrees, the slope of the relationship between maximum body and maximum air temperature is very close to one (Denny et al. 2006). Since the relationship between substratum, body, and air temperature is strong and approximately 1:1, the experimental warming of the substratum on black vs. white plates (2°C average) should increase body temperatures by approximately the same amount as would the global average change in air temperature expected by 2100 (IPCC 2013). One inevitable drawback of this experimental design is that not all species were affected equally. First, highly mobile organisms and dispersive life history stages are not affected by the manipulations except during the time they occupy the treatments. Second, not all species are heated by conduction to the same extent. Heated plates can only modify the immediate environment, and thus do not affect processes that occur at larger scales (larval survivorship, species pool) that will be influenced by climate change (Smale et al. 2011). The heated substratum would probably have little effect on more erect species such as mussels and erect algae (uncommon in these experiments), which have body temperatures more strongly determined by insolation, convection, and evaporation than by conduction (Bell 1995, Helmuth 1998). Nevertheless, the organisms found on these plates have body temperatures that are highly correlated to substratum temperatures (Denny and Harley 2006) Although using plates that varied in color was an effective method for manipulating temperature, some organisms may have been sensitive to the color or contrast of plate borders, independent of temperature. Previous studies have found that subtidal barnacle settlement and

146 recruitment appear to be higher on dark colored plates (Pomerat and Reiner 1942, Gregg 1945, James and Underwood 1994, Satheesh and Wesley 2010) although some studies have found similar settlement on different colors or rock types (Hurley 1973, Caffey 1982), and Gregg (1945) found that barnacle attachment was the same on surfaces with different black-white contrast. For intertidal barnacles, Lathlean and Minchinton (2012) found that settlement (measured daily) was higher on white plates than black plates, despite similar survivorship and recruitment on both colors. The differences that Lathlean and Minchinton observed may have been due to differences in biofilm development, and thus indirectly related to plate temperature. Although it seems that there is variability in barnacle color attraction, most studies to date have found that barnacles are more attracted to black or dark colors. If this is so, my estimates of barnacle abundance are conservative. Many mobile consumers (e.g. crabs and birds) search for prey visually, thus it is possible that different colored treatments may have affected consumption rates of organisms on my plates. I did not observe any crabs in surveys on the top surfaces of plates. In the ‘under plate’ surveys, crabs (Hemigrapsus sp.) comprised 1.7% of the total number of individual consumers and were found in equal quantities below white and black plates (ANOVA testing temperature treatment and zone: temperature: p > 0.25). Hermit crabs comprised <1% of total consumers and were also equally abundant under black and white plates (ANOVA testing temperature treatment and zone: temperature: p > 0.25). Therefore, I suspect that the influence of crabs on communities was low and similar for communities on both plates. In addition, plates were arranged on the shore in a haphazard but roughly alternating B-W-B-W fashion and were spaced 15-30 cm apart. Both black and white plates were noticeably different than the surrounding beige-orange bedrock with olive-brown seaweed (e.g., Fucus sp.). Therefore, I suspect that both plate colors would have been distinctive to birds and that birds would likely have indiscriminately eaten from neighboring plates of different colors, given their close proximity.

6.2.3 Relevance of solar plates to climate warming

While the experimental design used here represents an improvement on traditional field-based approaches, it too is an imperfect proxy for warmer future conditions. Climate change will alter the local environment more consistently, so all organisms will experience hotter temperatures. This is particularly important in the context of mobile consumers, whose feeding and handling

147 rates are influenced by temperature (Gillooly et al. 2001). In my experiments, consumers were in contact with the heated (or not) substratum while they were feeding, so they were likely influenced by temperature at that time. However, this and perhaps all, field based warming experiments are relatively open systems with respect to consumers (Williams 1994, Morelissen and Harley 2007, Post and Pedersen 2008, Wang et al. 2012, Williams et al. 2013). Results from manipulations of warming and trophic interactions may depend on the duration of heating experienced by consumers. In this regard, ‘natural experiments’ (Schiel et al. 2004, Woodward et al. 2010) and mesocosms (O'Connor 2009) may provide better insight regarding effects of warming on mobile organisms. As the climate warms, sea surface temperatures and aerial temperatures will rise, although air will warm more quickly than water (Sutton et al. 2007). The treatments used in this study were able to isolate the stressful effects of warming during aerial exposure (Hofmann and Somero 1995) from the generally non-stressful effects of changing seawater temperature (Menge et al. 2008). Temperature was the same across treatments during high tide, but these passively warmed plates differed by as much as 6°C between treatments during low tide, simulating an increase in stressful aerial conditions only. The disadvantage of this method is that the surrounding seawater could not also be warmed, which often has a beneficial effect on marine organisms. Warmer water temperatures in the temperate zone have been associated with increased development rates (Nasrolahi et al. 2011), growth rates (Menge et al. 2008), and feeding rates (Southward and Southward 1967, Sanford 2002, O'Connor 2009). The relative rate at which water and air warm and the potential benefits or stressors that result from each will determine the net thermal effect on intertidal organisms. Like most experimental approaches, these experiments also suffered from logistical constraints that limit their generality. There are three basic dimensions (time, space, and biological complexity) to balance with replication in an experimental design (Stewart et al. 2013). Because of necessary logistical trade-offs, there is no one perfect approach. The climate is changing at an unprecedented rate, however, organisms will have decades to adjust to warmer temperatures. The temporal scale of most experiments is much too short to extrapolate to many ecological (e.g., recovery) and evolutionary (e.g., adaptation) processes. Ecological processes (e.g., species interactions) track environmental conditions, which vary over spatial scales. Clearly the temporal and spatial scale of my experiments was too short and small to definitively

148 generalize to ecological communities of the future. In this dissertation, I chose to focus on capturing the effect of warming on higher levels of complexity, which meant trade-offs with respect to time and space. Despite the fact that these experiments were imperfect representations for warmer future conditions, my results demonstrate that thermal stress has significant impacts on intertidal communities. Field based experiments that increase thermal stress are still rare due to logistic difficulties, especially in marine systems (Wernberg et al. 2012a), but are more suitable for simulating future warming than many other field-based temperature manipulations.

6.3 Warming alters community dynamics via direct and indirect pathways

Warming affected organismal physiology and species interactions, which impacted community structure, diversity, and patterns of succession. Warming directly impacted two species of competing barnacles via reduced survivorship and growth rates (Chapter 4). This caused the space occupancy of barnacle populations to be lower, even after thermal stress subsided in winter. Both barnacle species were sensitive to warming at the organismal and population-level (Fig. 6.3, Box B). At the organismal level, survivorship and growth rates of both species were negatively affected by warming. This translated to reduced population density and space occupancy for both populations. However, because some B. glandula reproduction occurs after periods of lethal thermal stress (summer), the population density somewhat recovered after one year. Barnacles were major contributors to the difference in community structure between warm and cool treatments in summers (Chapters 3, 5). Barnacles were abundant in cool treatments but were less abundant and had smaller body sizes in warm treatments. Adult barnacles can facilitate other intertidal species by increasing topographic heterogeneity, thereby reducing desiccation and preventing consumption of spores and larvae (Menge 1976, Lubchenco 1983, Chapman 1989, van Tamelen et al. 1997, Barnes 2000, Harley 2006). In intertidal communities, many species of barnacles act as ‘habitat modifiers’; defined as an organism that significantly alters its local biotic or abiotic environment (Bruno and Bertness 2001). Warming directly negatively affected barnacle vital rates, which reduced barnacle populations, and resulted in altered community structure.

149 Figure 6.3: Direct and indirect effects of warming on key taxa in the intertidal zone of Salt Spring Island, BC. The taxa included represent those groups which consistently contributed to similarity in community structure. Direct (solid) and indirect (dashed) effects of warming (red arrows) and taxa (black arrows) on species abundances could be negative (-) or positive (+). Competitive interactions via preemption of space (black lines ending in solid circles) in the intertidal are inferred from relative abundances in my experiments and / or previous studies. Thick black-red lines indicate modification (m) of an interaction; e.g., warming modifies the effect of limpets on B. glandula. Box A: displays how warming and limpets interact to influence emergent properties (EP) of the community; warming and limpets independently affect diversity, while limpets and warming interact to affect community structure and succession. Box B: details the direct effect of warming on vital rates and life stages of barnacle species, Balanus glandula (green) and Chthamalus dalli (purple). Grey arrows indicate life stage transitions and vital rates. Stages or processes affected by temperature are indicated in red text. Responses quantified in other studies are denoted by superscripts1. Underlined processes were measured in warmed water, while the remaining effects were quantified in warmed aerial conditions. Processes in black text have not been experimentally tested.

1 Berger and Emlet 2007; 2 Southward and Southward 1967; 3 Emlet 2006; 4 Barnes and Barnes 1956; 5 Miller et al. 1989

150 Warming also indirectly affected community dynamics via altered species interactions. Warming strengthened the indirect positive effect of limpets on B. glandula and modified the effect of limpet consumption on diatoms (Fig. 6.3). Since B. glandula and diatoms were two of the three species with the highest contribution to dissimilarity between warm and cool treatments in summers (Chapter 5), community structure and succession may have been modified indirectly via these interactions. Although the exact mechanisms responsible for the warming induced change in interactions was not explicitly tested in these experiments, it is likely that interactions with tertiary taxa were important. Species interactions vary in space, time, ontogeny, and over environmental gradients (Callaway and Walker 1997, Barton 2010, Kordas and Dudgeon 2011), but can also be modified by other species (Barton and Schmitz 2009). The experiments within my dissertation demonstrate that warming acts directly and indirectly to determine emergent properties of communities. It is no surprise that the impact of warming on communities is complex. However, by uncovering the various pathways affected by temperature, we can prepare for and respond to the effects of global environmental change.

6.4 Looking forward

Knowing that communities will be affected by warming via both direct and indirect pathways emphasizes the need to better understand each of these processes. Considerable research has examined the causes and consequences of responses of organisms, populations, and species interactions to warming, but in some ways, the field of climate change ecology is still in its infancy. There remains a limited ability to predict the outcomes of perturbations to communities and ecosystems. Here (and in Chapter 2) I discuss some priorities for future research to enhance our understanding of future warming on ecological communities.

6.4.1 Global warming will be stressful

In the face of rising global temperatures, habitat temperatures will approach and in many cases exceed the thermal optima for resident species. Organism performance is positively correlated with temperature, up to an optimum, after which it declines sharply. The rising portion of this thermal performance curve (TPC) has been the focus of metabolic theory for decades. However, the declining portion of the TPC has been largely ignored by terrestrial and freshwater ecologists

151 because the biological processes that lead to performance declines are completely different from those that convey benefits. The body temperatures of most ectotherms match or exceed their physiological thermal limits when in exposed habitats (Sunday et al. 2014). Thus, rising mean temperature and increasing frequency and duration of thermal events will likely push habitat temperatures past thermal optima (Vasseur et al. 2014). Thus, it is essential to understand the declining portion of these curves. TPCs are a useful currency for understanding organismal responses to changing temperature. Indeed a library of TPC information has been compiled by researchers in Germany (Dell et al. 2011, 2013), providing a valuable starting point for assessing organismal responses to stress. However, given that the shape of the TPC varies widely among species, life stage, media (e.g., water, soil, air), and the response being measured, additional studies are needed.

6.4.2 Thermal dependency of species interactions

Comparing interspecific TPCs is a useful null model to predict how species interactions will respond to warming. Interacting species may have thermal performance curves that differ in the magnitude of the response (Fig. 6.4a), the position of peak performance relative to temperature (Fig. 6.4b), or in the rate of response (Fig. 6.4c-f). The effect of warming on the interaction will depend on each species thermal performance curve, and the location (i.e., initial-final temperature) and magnitude of temperature change along the x-axis. There are several ways that warming can modify interactions between two species. In general, (1) warming can act in opposing directions, increasing the performance of one species while lowering the performance of the other (Fig. 6.4b; Wethey 1984a, Grigaltchik et al. 2012, Fey and Cottingham 2012), which would switch the sign of an interaction. (2) warming can be differentially beneficial for two species (Sanford 1999, O'Connor 2009), which could weaken (Fig. 6.4c) or strengthen (Fig. 6.4d) the interaction depending on the shape of their performance curves. Or (3) differentially detrimental for two species (Poore et al. 2013), which could also weaken (Fig. 6.4e) or strengthen (Fig. 6.4f) the interaction depending on the relative shape of the curves. Because individual thermal performance responses are unimodal and are often asymmetrical between pairs (Dell et al. 2011), the state of the interaction (above) will depend on the magnitude of warming. Thus interactions can change in sign or strength as temperatures warm or cool.

152 Additional studies that explicitly quantify TPCs for interacting pairs would be useful in describing how interactions will change in a warmer future.

Figure 6.4: The thermal dependency of species interactions. Thermal performance curves (TPCs) can vary in the trait performance value (a) or temperature at which the curve peaks (b) and in the rate of response (c-f). The change in the interaction will depend on the relative shape of the curves as well as change in temperature from x to y. Even if the change in temperature is the same (as shown here), the starting conditions relative to the peaks will determine if interactions weaken (c, e) or strengthen (d, f).

153 6.4.3 Community and food web responses

Most of what we know about community and food web responses to warming come from metabolic theory rather than from communities under stress. Metabolic theory predicts that as temperature increases, the metabolic rates of all species increase, resulting in an increasing demand for and consumption of energy at each trophic level. When the supply of energy transferred up the food chain is no longer sufficient to support the minimum viable population size of the top predator, that species is lost. There is no corresponding prediction or theory for communities in stressful conditions. Collecting TPCs for an entire community may serve as a solid starting point or null model to describe the effect of rising temperatures. In a community context, there are likely to be unexpected feedbacks and emergent effects that would not be described by this simple model, but it would serve as a jumping off point for further study. A significant fraction of anthropogenic climate change is irreversible on a multi-century to millennial time scale (IPCC 2013). Once CO2 emissions have ceased, surface temperatures will remain approximately constant at elevated levels for many centuries (IPCC 2013). As organisms struggle to acclimate and adapt to rapidly changing conditions, so too are ecologists struggling to understand these changes with the hope of predicting what may come. To enhance the accuracy of these predictions, it is vital to understand how ecological communities are structured; with organisms directly affected by their abiotic environment and indirectly affected by neighboring individuals. Understanding these (not so) basic pathways and their thermal dependencies will pave the way for predictive climate change ecology. Historically, the focus has been on abiotic drivers of change. In the last decade, the importance of biotic influences has been repeatedly stressed. I would now add that the interdependency of these two influences need also be considered in the context of communities in a warmer world.

154 Bibliography

Admiraal, W. 1977. Influence of light and temperature on the growth rate of estuarine benthic diatoms in culture. Marine Biology 39:1–9. Admiraal, W., and H. Peletier. 1980. Influence of seasonal variations of temperature and light on the growth rate of cultures and natural populations of intertidal diatoms. Marine Ecology Progress Series 2:35–43. Allison, G. 2004. The influence of species diversity and stress intensity on community resistance and resilience. Ecological Monographs 74:117–134. Alsterberg, C., J. S. Eklöf, L. Gamfeldt, J. N. Havenhand, and K. Sundbäck. 2013. Consumers mediate the effects of experimental ocean acidification and warming on primary producers. Proceedings of the National Academy of Sciences 110:8603–8608. Anderson, M., R. N. Gorley, and R. K. Clarke. 2008. Permanova+ for Primer: Guide to Software and Statistical Methods. Plymouth, UK. Assessment, M. E. 2005. Ecosystems and human well-being: synthesis. Pages 1–155. Island Press, Washington, DC. Barbraud, C., P. Rivalan, P. Inchausti, M. Nevoux, V. Rolland, and H. Weimerskirch. 2010. Contrasted demographic responses facing future climate change in Southern Ocean seabirds. Journal of Animal Ecology 80:89–100. Barnes, H. 1955. The growth rate of Balanus balanoides (L.). Oikos 6:109–113. Barnes, H., and M. Barnes. 1956. The general biology of Balanus glandula Darwin. Pacific Science 10:415–422. Barnes, M. 2000. The use of intertidal barnacle shells. Oceanography and Marine Biology: an Annual Review 38:157–187. Barnosky, A. D., E. A. Hadly, J. Bascompte, E. L. Berlow, J. H. Brown, M. Fortelius, W. M. Getz, J. Harte, A. Hastings, P. A. Marquet, N. D. Martinez, A. Mooers, P. Roopnarine, G. Vermeij, J. W. Williams, R. Gillespie, J. Kitzes, C. Marshall, N. Matzke, D. P. Mindell, E. Revilla, and A. B. Smith. 2012. Approaching a state shift in Earth's biosphere. Nature 486:52–58. Barry, J. P., C. H. Baxter, R. D. Sagarin, and S. Gilman. 1995. Climate-related, long-term faunal

155 changes in a California rocky intertidal community. Science 267:672–675. Barton, B. T. 2010. Climate warming and predation risk during herbivore ontogeny. Ecology 91:2811–2818. Barton, B. T., and O. J. Schmitz. 2009. Experimental warming transforms multiple predator effects in a grassland food web. Ecology Letters 12:1317–1325. Bates, A. E., B. J. Hilton, and C. D. G. Harley. 2009. Effects of temperature, season and locality on wasting disease in the keystone predatory sea star Pisaster ochraceus. Diseases of Aquatic Organisms 86:245–251. Bates, D. M., M. Maechler, and B. Bolker. 2011. lme4: Linear mixed-effects models using S4 classes. R Foundation for Statistical Computing. Beaugrand, G., P. C. Reid, F. Ibañez, J. A. Lindley, and M. Edwards. 2002. Reorganization of North Atlantic marine copepod biodiversity and climate. Science 296:1692–1694. Bell, E. C. 1993. Photosynthetic response to temperature and desiccation of the intertidal alga Mastocarpus papillatus. Marine Biology 117:337–346. Bell, E. C. 1995. Environmental and morphological influences on thallus temperature and desiccation of the intertidal alga Mastocarpus papillatus Kützing. Journal of Experimental Marine Biology and Ecology 191:29–55. Berger, M. S., and R. B. Emlet. 2007. Heat-shock response of the upper intertidal barnacle Balanus glandula: Thermal stress and acclimation. Biological Bulletin 212:232–241. Berlow, E., and S. Navarrete. 1997. Spatial and temporal variation in rocky intertidal community organization: Lessons from repeating field experiments. Journal of Experimental Marine Biology and Ecology 214:195–229. Bertness, M. D. 1989. Intraspecific competition and facilitation in a northern acorn barnacle population. Ecology 70:257–268. Bertness, M. D., and G. H. Leonard. 1997. The role of positive interactions in communities: lessons from intertidal habitats. Ecology 78:1976–1989. Bertness, M. D., G. H. Leonard, J. M. Levine, and J. F. Bruno. 1999a. Climate-driven interactions among rocky intertidal organisms caught between a rock and a hot place. Oecologia 120:446–450. Bertness, M. D., G. H. Leonard, J. M. Levine, P. R. Schmidt, and A. O. Ingraham. 1999b. Testing the relative contribution of positive and negative interactions in rocky intertidal

156 communities. Ecology 80:2711–2726. Bertness, M. D., S. Gaines, and S. Yeh. 1998. Making mountains out of barnacles: The dynamics of acorn barnacle hummocking. Ecology 79:1382–1394. Bischoff, B., and C. Wiencke. 1995a. Temperature adaptation in strains of the amphi-equatorial green alga Urospora penicilliformis (Acrosiphoniales): biogeographical implications. Marine Biology 122:681–688. Bischoff, B., and C. Wiencke. 1995b. Temperature ecotypes and biogeography of Acrosiphoniales (Chlorophyta) with Arctic-Antarctic disjunct and Arctic/cold-temperature distributions. European Journal of Phycology 30:19–27. Blanchette, C. A., M. C. Miner, P. T. Raimondi, D. Lohse, K. E. K. Heady, and B. R. Broitman. 2008. Biogeographical patterns of rocky intertidal communities along the Pacific coast of North America. Journal of Biogeography 35:1593–1607. Bradshaw, W., and C. Holzapfel. 2006. Evolutionary response to rapid climate change. Science 312:1477–1478. Branch, G. M. 1981. The biology of limpets: physical factors, energy flow, and ecological interactions. Oceanography and Marine Biology: an Annual Review 19:235–380. Broitman, B. R., C. A. Blanchette, B. A. Menge, J. Lubchenco, C. Krenz, M. Foley, P. T. Raimondi, D. Lohse, and S. D. Gaines. 2008. Spatial and temporal patterns of invertebrate recruitment along the west coast of the United States. Ecological Monographs 78:403–421. Bruno, J. F., and M. D. Bertness. 2001. Habitat modification and facilitation in benthic marine communities. Pages 201–218 in M. D. Bertness, M. E. Hay, and S. D. Gaines, editors. Marine Community Ecology. Sinauer, Sunderland, MA. Bruno, J., J. Stachowicz, and M. Bertness. 2003. Inclusion of facilitation into ecological theory. Trends in Ecology & Evolution 18:119–125. Buckley, L. B., S. A. Waaser, H. J. MacLean, and R. Fox. 2011. Does including physiology improve species distribution model predictions of responses to recent climate change? Ecology 92:2214–2221. Burgmer, T., and H. Hillebrand. 2010. Temperature mean and variance alter phytoplankton biomass and biodiversity in a long-term microcosm experiment. Oikos 120:922–933. Burnaford, J. L. 2004. Habitat modification and refuge from sublethal stress drive a marine plant-herbivore association. Ecology 85:2837–2849.

157 Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Second edition. Springer-Verlag, New York. Caffey, H. M. 1982. No effect of naturally-occurring rock types on settlement or survival in the intertidal barnacle, Tesseropora rosea (Krauss). Journal of Experimental Marine Biology and Ecology 63:119–132. Callaway, R. M., and L. R. Walker. 1997. Competition and facilitation: a synthetic approach to interactions in plant communities. Ecology 78:1958–1965. Campbell, M., and S. Farrell. 2006. Biochemistry. Carscadden, J., B. Nakashima, and K. Frank. 1997. Effects of fish length and temperature on the timing of peak spawning in capelin (Mallotus villosus). Canadian Journal Of Fisheries And Aquatic Sciences 54:781–787. Castenholz, R. W. 1961. The effect of grazing on marine littoral diatom populations. Ecology 42:783–794. Cáceres-Puig, J. I., and F. Abasolo-Pacheco. 2007. Effect of temperature on growth and survival of Crassostrea corteziensis spat during late-nursery culturing at the hatchery. Aquaculture 272:417-422. Chapman, A. 1989. Abundance of Fucus spiralis and ephemeral seaweeds in a high eulittoral zone: effects of grazers, canopy and substratum type. Marine Biology 102:565–572. Cheung, W. W., V. W. Lam, J. L. Sarmiento, K. Kearney, R. Watson, and D. Pauly. 2009. Projecting global marine biodiversity impacts under climate change scenarios. Fish and Fisheries 10:235–251. Claireaux, G., and C. Lefrançois. 2007. Linking environmental variability and fish performance: integration through the concept of scope for activity. Philosophical Transactions of the Royal Society B: Biological Sciences 362:2031–2041. Clarke, K. R., and R. M. Warwick. 2001. Change in marine communities: An approach to statistical analysis and interpretation. PRIMER-E. United Kingdom, Plymouth. Clarke, K. R., and R. N. Gorley. 2006. PRIMER v6: User manual / tutorial. PRIMER-E Ltd, Plymouth, UK. Clarke, K. R., P. J. Somerfield, L. Airoldi, and R. M. Warwick. 2006. Exploring interactions by second-stage community analyses. Journal of Experimental Marine Biology and Ecology 338:179–192.

158 Connell, J. 1961a. The influence of interspecific competition and other factors on the distribution of the barnacle Chthamalus stellatus. Ecology 42:710–723. Connell, J. 1961b. Effects of competition, predation by Thais lapillus, and other factors on natural populations of the barnacle Balanus balanoides. Ecological Monographs 31:61–104. Coquelle, N., E. Fioravanti, M. Weik, and F. Vellieux. 2007. Activity, stability and structural studies of lactate dehydrogenases adapted to extreme thermal environments. Journal of Molecular Biology 374:547–562. Costanza, R., R. dArge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S. Naeem, R. V. O'Neill, J. Paruelo, R. Raskin, P. Sutton, and M. van den Belt. 1997. The value of the world's ecosystem services and natural capital. Nature 387:253–260. Costello, J. H., B. K. Sullivan, and D. J. Gifford. 2006. A physical-biological interaction underlying variable phenological responses to climate change by coastal zooplankton. Journal of Plankton Research 28:1099–1105. Cox, T. E., and C. M. Smith. 2011. Thermal ecology on an exposed algal reef: infrared imagery a rapid tool to survey temperature at local spatial scales. Coral Reefs 30:1109–1120. Crain, C. M., K. Kroeker, and B. S. Halpern. 2008. Interactive and cumulative effects of multiple human stressors in marine systems. Ecology Letters 11:1304–1315. Crisp, D. J., and D. A. Ritz. 1967. Changes in temperature tolerance of Balanus balanoides during its life-cycle. Helgoländer wissenschaftliche Meeresuntersuchungen 15:98–115. Cushing, D. H. 1982. Climate and fisheries. Academic Press, London. Dahlhoff, E. P., J. H. Stillman, and B. A. Menge. 2002. Physiological community ecology: Variation in metabolic activity of ecologically important rocky intertidal invertebrates along environmental gradients. Integrative and Comparative Biology 42:862–871. Dalgleish, H. J., D. N. Koons, M. B. Hooten, C. A. Moffet, and P. B. Adler. 2011. Climate influences the demography of three dominant sagebrush steppe plants. Ecology 92:75–85. Darwin, C. 1859. On the Origin of Species. John Murray, United Kingdom. Daufresne, M., K. Lengfellner, and U. Sommer. 2009. Global warming benefits the small in aquatic ecosystems. Proceedings of the National Academy of Sciences 106:12788–12793. Davenport, J., and J. Davenport. 2005. Effects of shore height, wave exposure and geographical distance on thermal niche width of intertidal fauna. Marine Ecology Progress Series 292:41– 50.

159 Davis, A. J., J. H. Lawton, B. Shorrocks, and L. S. Jenkinson. 2002. Individualistic species responses invalidate simple physiological models of community dynamics under global environmental change. Journal of Animal Ecology 67:600–612. Davis, A. J., L. S. Jenkinson, J. H. Lawton, B. Shorrocks, and S. Wood. 1998. Making mistakes when predicting shifts in species range in response to global warming. Nature 391:783–786. Davison, I. R., and G. A. Pearson. 2008. Stress tolerance in intertidal seaweeds. Journal of Phycology 32:197–211. Dayton, P. 1971. Competition, disturbance, and community organization: provision and subsequent utilization of space in a rocky intertidal community. Ecological Monographs 41:351–389. Delaney, M. P. 2003. Effects of temperature and turbulence on the predator-prey interactions between a heterotrophic flagellate and a marine bacterium. Microbial Ecology 45:218–225. Dell, A. I., S. Pawar, and V. M. Savage. 2011. Systematic variation in the temperature dependence of physiological and ecological traits. Proceedings of the National Academy of Sciences 108:10591–10596. Dell, A. I., S. Pawar, and V. M. Savage. 2013. Temperature dependence of trophic interactions are driven by asymmetry of species responses and foraging strategy. Journal of Animal Ecology 83:70–84. Demes, K. W., R. L. Kordas, and J. P. Jorve. 2012. Ferry wakes increase seaweed richness and abundance in a sheltered rocky intertidal habitat. Hydrobiologia 693:1–11. Denny, M. W., and C. D. G. Harley. 2006. Hot limpets: predicting body temperature in a conductance-mediated thermal system. Journal of Experimental Biology 209:2409–2419. Denny, M. W., L. P. Miller, and C. D. G. Harley. 2006. Thermal stress on intertidal limpets: long-term hindcasts and lethal limits. Journal of Experimental Biology 209:2420–2431. Dethier, M. N., and D. O. Duggins. 1984. An“ indirect commensalism” between marine herbivores and the importance of competitive hierarchies. American Naturalist 124:205–219. Dethier, M., S. Williams, and A. Freeman. 2005. Seaweeds under stress: Manipulated stress and herbivory affect critical life-history functions. Ecological Monographs 75:403–418. Doney, S. C., M. Ruckelshaus, J. E. Duffy, J. P. Barry, F. Chan, C. A. English, H. M. Galindo, J. M. Grebmeier, A. B. Hollowed, N. Knowlton, J. Polovina, N. N. Rabalais, W. J. Sydeman, and L. D. Talley. 2012. Climate change impacts on marine ecosystems. Annual Review of

160 Marine Science 4:11–37. Dong, Y., L. P. Miller, J. G. Sanders, and G. N. Somero. 2008. Heat-shock protein 70 (Hsp70) expression in four limpets of the genus Lottia: interspecific variation in constitutive and inducible synthesis correlates with in situ exposure to heat stress. Biological Bulletin 215:173–181. Dunson, W. A., and J. Travis. 1991. The role of abiotic factors in community organization. American Naturalist 138:1067–1091. Easterling, D. R. 2000. Climate extremes: Observations, modeling, and impacts. Science 289:2068–2074. Edwards, M., and A. J. Richardson. 2004. Impact of climate change on marine pelagic phenology and trophic mismatch. Nature 430:881–884. Eklöf, J. S., C. Alsterberg, J. N. Havenhand, K. Sundbäck, H. L. Wood, and L. Gamfeldt. 2012. Experimental climate change weakens the insurance effect of biodiversity. Ecology Letters 15:864–872. Elton, C. 1927. Animal Ecology. Sidgwick and Jackson, London. Elton, C. S., and R. S. Miller. 1954. The ecological survey of animal communities: with a practical system of classifying habitats by structural characters. The Journal of Ecology 42:460–496. Emlet, R. B. 2006. Linking stages of life history: How larval quality translates into juvenile performance for an intertidal barnacle (Balanus glandula). Integrative and Comparative Biology 46:334–346. Farrell, T. M. 1991. Models and mechanisms of succession: an example from a rocky intertidal community. Ecological Monographs 61:95–113. Fey, S. B., and K. L. Cottingham. 2012. Thermal sensitivity predicts the establishment success of nonnative species in a mesocosm warming experiment. Ecology 93:2313–2320. Finlay, J. A., B. R. Fletcher, M. E. Callow, and J. A. Callow. 2008. Effect of background colour on growth and adhesion strength of Ulva sporelings. Biofouling 24:219–225. Firth, L. B., T. P. Crowe, P. Moore, R. C. Thompson, and S. J. Hawkins. 2009. Predicting impacts of climate-induced range expansion: an experimental framework and a test involving key grazers on temperate rocky shores. Global Change Biology 15:1413–1422. Foster, B. A. 1969. Tolerance of high temperatures by some intertidal barnacles. Marine Biology

161 4:326–332. Foster, B. A. 1971. On the determinants of the upper limit of intertidal distribution of barnacles (Crustacea: Cirripedia). The Journal of Animal Ecology 40:33–48. Freitas, V., J. Campos, M. Fonds, and H. Vanderveer. 2007. Potential impact of temperature change on epibenthic predator–bivalve prey interactions in temperate estuaries. Journal of Thermal Biology 32:328–340. Gattuso, J. P., J. Hinkel, A. Khattabi, K. McInnes, Y. Saito, and A. Sallenger. 2014. Coastal Systems and Low-Lying Areas. Pages 1–85 in IPCC WGII AR5. Gedan, K. B., and M. D. Bertness. 2009. Experimental warming causes rapid loss of plant diversity in New England salt marshes. Ecology Letters 12:842–848. Gedan, K. B., J. Bernhardt, M. D. Bertness, and H. M. Leslie. 2011. Substrate size mediates thermal stress in the rocky intertidal. Ecology 92:576–582. Gilg, O., K. M. Kovacs, J. Aars, J. Fort, G. Gauthier, D. Gremillet, R. A. Ims, H. Meltofte, J. Moreau, E. Post, N. M. Schmidt, G. Yannic, and L. Bollache. 2012. Climate change and the ecology and evolution of Arctic vertebrates. Annals of the New York Academy of Sciences 1249:166–190. Gillooly, J. F., E. L. Charnov, G. B. West, V. M. Savage, and J. H. Brown. 2002. Effects of size and temperature on developmental time. Nature 417:70–73. Gillooly, J., J. Brown, G. West, V. Savage, and E. Charnov. 2001. Effects of size and temperature on metabolic rate. Science 293:2248–2251. Gosselin, L. A., and P.-Y. Qian. 1996. Early post-settlement mortality of an intertidal barnacle: a critical period for survival. Marine Ecology Progress Series 135:69–75. Graham, R. M., and E. C. Grimm. 1990. Effects of global climate change on the patterns of terrestrial biological communities. Trends in Ecology & Evolution 5:289–292. Graham, R. M., E. Lundelius, M. Graham, E. Schroeder, R. Toomey, E. Anderson, A. Barnosky, J. Burns, C. Churcher, D. Grayson, R. Guthrie, C. Harington, G. Jefferson, L. Martin, H. McDonald, R. Morlan, H. Semken, S. Webb, L. Werdelin, and M. Wilson. 1996. Spatial response of mammals to late quaternary environmental fluctuations. Science 272:1601– 1606. Gregg, J. H. 1945. Background illumination as a factor in the attachment of barnacle cyprids. Biological Bulletin 88:44–49.

162 Grigaltchik, V. S., A. J. W. Ward, and F. Seebacher. 2012. Thermal acclimation of interactions: differential responses to temperature change alter predator-prey relationship. Proceedings of the Royal Society of London B: Biological Sciences 279:4058–4064. Grosholz, E. D., G. M. Ruiz, C. A. Dean, K. A. Shirley, and J. L. Maron. 2000. The impacts of a nonindigenous marine predator in a California bay. Ecology 81:1206–1224. Hairston, N. G. 1989. Ecological experiments: purpose, design and execution. Cambridge University Press Cambridge. Hallegraeff, G. M. 2010. Ocean climate change, phytoplankton community responses, and harmful algal blooms: a formidable predictive challenge. Journal of Phycology 46:220–235. Halpern, B. S., S. Walbridge, K. A. Selkoe, C. V. Kappel, F. Micheli, C. D'Agrosa, J. F. Bruno, K. S. Casey, C. Ebert, H. E. Fox, R. Fujita, D. Heinemann, H. S. Lenihan, E. M. P. Madin, M. T. Perry, E. R. Selig, M. Spalding, R. Steneck, and R. Watson. 2008. A global map of human impact on marine ecosystems. Science 319:948–952. Harley, C. D. G. 2002. Light availability indirectly limits herbivore growth and abundance in a high rocky intertidal community during the winter. Limnology and Oceanography 47:1217– 1222. Harley, C. D. G. 2003. Abiotic stress and herbivory interact to set range limits across a two- dimensional stress gradient. Ecology 84:1477–1488. Harley, C. D. G. 2006. Effects of physical ecosystem engineering and herbivory on intertidal community structure. Marine Ecology Progress Series 317:29–39. Harley, C. D. G. 2011. Climate change, keystone predation, and biodiversity loss. Science 334:1124–1127. Harley, C. D. G. 2013. Linking ecomechanics and ecophysiology to interspecific interactions and community dynamics. Annals of the New York Academy of Sciences 1297.1:73–82. Harley, C. D. G., A. Hughes, K. Hultgren, B. Miner, C. Sorte, C. Thornber, L. Rodriguez, L. Tomanek, and S. Williams. 2006. The impacts of climate change in coastal marine systems. Ecology Letters 9:228–241. Harley, C. D. G., and B. S. T. Helmuth. 2003. Local- and regional-scale effects of wave exposure, thermal stress, and absolute versus effective shore level on patterns of intertidal zonation. Limnology and Oceanography 48:1498–1508. Harley, C. D. G., and J. L. O’Riley. 2011. Non-linear density-dependent effects of an intertidal

163 ecosystem engineer. Oecologia 166:531–541. Harley, C. D. G., and J. P. Lopez. 2003. The natural history, thermal physiology, and ecological impacts of intertidal mesopredators, Oedoparena spp. (Diptera : Dryomyzidae). Invertebrate Biology 122:61–73. Harley, C. D. G., and R. T. Paine. 2009. Contingencies and compounded rare perturbations dictate sudden distributional shifts during periods of gradual climate change. Proceedings of the National Academy of Sciences 106:11172–11176. Harte, J., and R. Shaw. 1995. Shifting dominance within a montane vegetation community: results of a climate-warming experiment. Science 267:876–880. Harvell, C. D., C. E. Mitchell, J. R. Ward, S. Altizer, A. P. Dobson, R. S. Ostfeld, and M. D. Samuel. 2002. Climate warming and disease risks for terrestrial and marine biota. Science 296:2158–2162. Hawkins, S. J., H. E. Sugden, N. Mieszkowska, P. J. Moore, E. Poloczanska, R. Leaper, R. J. H. Herbert, M. J. Genner, P. S. Moschella, R. C. Thompson, S. R. Jenkins, A. J. Southward, and M. T. Burrows. 2009. Consequences of climate-driven biodiversity changes for ecosystem functioning of North European rocky shores. Marine Ecology Progress Series 396:245–259. Hawkins, S. J., P. J. Moore, M. T. Burrows, E. Poloczanska, N. Mieszkowska, R. J. H. Herbert, S. R. Jenkins, R. C. Thompson, M. J. Genner, and A. J. Southward. 2008. Complex interactions in a rapidly changing world: responses of rocky shore communities to recent climate change. Climate Research 37:123–133. Hays, G. C., A. J. Richardson, and C. Robinson. 2005. Climate change and marine plankton. Trends in Ecology & Evolution 20:337–344. Helmuth, B. 1998. Intertidal mussel microclimates: predicting the body temperature of a sessile invertebrate. Ecological Monographs 68:51–74. Helmuth, B. 2009. From cells to coastlines: how can we use physiology to forecast the impacts of climate change? Journal of Experimental Biology 212:753–760. Helmuth, B., B. R. Broitman, C. A. Blanchette, S. Gilman, P. Halpin, C. D. G. Harley, M. J. O'Donnell, G. E. Hofmann, B. Menge, and D. Strickland. 2006a. Mosaic patterns of thermal stress in the rocky intertidal zone: implications for climate change. Ecological Monographs 76:461–479. Helmuth, B., C. D. G. Harley, P. M. Halpin, M. O'Donnell, G. E. Hofmann, and C. A.

164 Blanchette. 2002. Climate change and latitudinal patterns of intertidal thermal stress. Science 298:1015–1017. Helmuth, B., J. Kingsolver, and E. Carrington. 2005. Biophysics, physiological ecology, and climate change: Does mechanism matter? Annual Review of Physiology 67:177–201. Helmuth, B., N. Mieszkowska, P. Moore, and S. J. Hawkins. 2006b. Living on the edge of two changing worlds: Forecasting the responses of rocky intertidal ecosystems to climate change. Annual Review of Ecology Evolution and Systematics 37:373–404. Hewatt, W. G. 1935. Ecological succession in the Mytilus californianus habitat as observed in Monterey Bay, California. Ecology 16:244–251. Hines, A. H. 1978. Reproduction in three species of intertidal barnacles from central California. Biological Bulletin 154:262–281. Hochachka, P., and G. Somero. 2002. Biochemical adaptation: mechanism and process in physiological evolution. Pages 1–466. Oxford University Press. Hoegh-Guldberg, O., and J. F. Bruno. 2010. The impact of climate change on the world's marine ecosystems. Science 328:1523–1528. Hofmann, G. E., and G. N. Somero. 1995. Evidence for protein damage at environmental temperatures: seasonal changes in levels of ubiquitin conjugates and hsp70 in the intertidal mussel Mytilus trossulus. Journal of Experimental Biology 198:1509–1518. Hsiao, S. I. 1970. Light and temperature effects on the growth, morphology, and reproduction of Petalonia fascia. Canadian Journal of Botany 48:1359–1361. Huey, R. B., M. R. Kearney, A. Krockenberger, J. A. M. Holtum, M. Jess, and S. E. Williams. 2012. Predicting organismal vulnerability to climate warming: roles of behaviour, physiology and adaptation. Philosophical Transactions of the Royal Society B: Biological Sciences 367:1665–1679. Huey, R., and D. Berrigan. 2001. Temperature, demography, and ectotherm fitness. American Naturalist 158:204–210. Hughes, L. 2000. Biological consequences of global warming: is the signal already apparent? Trends in Ecology & Evolution 15:56–61. Hurley, A. C. 1973. Larval settling behaviour of the acorn barnacle (Balanus pacificus Pilsbry) and its relation to distribution. The Journal of Animal Ecology 42:599–609. Hutchins, L. W. 1947. The bases for temperature zonation in geographical distribution.

165 Ecological Monographs 17:325–335. Ings, T. C., J. M. Montoya, J. Bascompte, N. Blüthgen, L. Brown, C. F. Dormann, F. Edwards, D. Figueroa, U. Jacob, J. I. Jones, R. B. Lauridsen, M. E. Ledger, H. M. Lewis, J. M. Olesen, F. F. Van Veen, P. H. Warren, and G. Woodward. 2009. Ecological networks - beyond food webs. Journal of Animal Ecology 78:253–269. IPCC. 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Pages 1–104 (R. K. Pachauri and A. Reisinger, Eds.). Geneva, Switzerland. IPCC. 2013. Summary for Policymakers. In: Climate Change 2013: The Physical Science Basis. Working Group I Contribution to the IPCC Fifth Assessment Report. Pages 1–36 (T. F. Stocker, D. Quin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P. M. Midgley, Eds.). Cambridge University Press, Cambridge, UK. James, R. J., and A. J. Underwood. 1994. Influence of colour of substratum on recruitment of spirorbid tubeworms to different types of intertidal boulders. Journal of Experimental Marine Biology and Ecology 181:105–115. Johnson, C. R., S. C. Banks, N. S. Barrett, F. Cazassus, P. K. Dunstan, G. J. Edgar, S. D. Frusher, C. Gardner, M. Haddon, F. Helidoniotis, K. L. Hill, N. J. Holbrook, G. W. Hosie, P. R. Last, S. D. Ling, J. Melbourne-Thomas, K. Miller, G. T. Pecl, A. J. Richardson, K. R. Ridgway, S. R. Rintoul, D. A. Ritz, D. J. Ross, J. C. Sanderson, S. A. Shepherd, A. Slotwinski, K. M. Swadling, and N. Taw. 2011. Climate change cascades: Shifts in oceanography, species' ranges and subtidal marine community dynamics in eastern Tasmania. Journal of Experimental Marine Biology and Ecology 400:17–32. Johnson, L. E. 1992. Potential and peril of field experimentation: the use of copper to manipulate molluscan herbivores. Journal of Experimental Marine Biology and Ecology 160:251–262. Jones, A. M., R. Berkelmans, M. J. H. van Oppen, J. C. Mieog, and W. Sinclair. 2008. A community change in the algal endosymbionts of a scleractinian coral following a natural bleaching event: field evidence of acclimatization. Proceedings of the Royal Society of London B: Biological Sciences 275:1359–1365. Jonsson, B., and N. Jonsson. 2009. A review of the likely effects of climate change on anadromous Atlantic salmon Salmo salar and brown trout Salmo trutta, with particular reference to water temperature and flow. Journal of Fish Biology 75:2381–2447.

166 Kavanaugh, M. T., K. J. Nielsen, F. T. Chan, B. A. Menge, R. M. Letelier, and L. M. Goodrich. 2009. Experimental assessment of the effects of shade on an intertidal kelp: Do phytoplankton blooms inhibit growth of open coast macroalgae? Limnology and Oceanography 54:276–288. Kim, K. Y., and D. J. Garbary. 2006. Fluorescence responses of photosynthesis to extremes of hyposalinity, freezing and desiccation in the intertidal crust Hildenbrandia rubra (Hildenbrandiales, Rhodophyta). Phycologia 45:680–686. Kitamura, A. 2004. Effects of seasonality, forced by orbital-insolation cycles, on offshore molluscan faunal change during rapid warming in the Sea of Japan. Palaeogeography, Palaeoclimatology, Palaeoecology 203:169–178. Kordas, R. L., and S. Dudgeon. 2011. Dynamics of species interaction strength in space, time and with developmental stage. Proceedings of the Royal Society of London B: Biological Sciences 278:1804–1813. Kordas, R. L., C. D. G. Harley, and M. I. O'Connor. 2011. Community ecology in a warming world: the influence of temperature on interspecific interactions in marine systems. Journal of Experimental Marine Biology and Ecology 400:218–226. Kudo, I., M. Miyamoto, Y. Noiri, and Y. Maita. 2000. Combined effects of temperature and iron on the growth and physiology of the marine diatom Phaeodactylum tricornutum (Bacillariophyceae). Journal of Phycology 36:1096–1102. Kuznetsova, A., P. B. Brockhoff, and R. H. B. Christensen. 2012. Tests for random and fixed effects for linear mixed effect models (lmer objects of lme4 package). Lang, B., B. C. Rall, and U. Brose. 2011. Warming effects on consumption and intraspecific interference competition depend on predator metabolism. Journal of Animal Ecology 81:516–523. Lathlean, J. A., and T. E. Minchinton. 2012. Manipulating thermal stress on rocky shores to predict patterns of recruitment of marine invertebrates under a changing climate. Marine Ecology Progress Series 467:121–136. Lathlean, J. A., D. J. Ayre, and T. E. Minchinton. 2013. Temperature variability at the larval scale affects early survival and growth of an intertidal barnacle. Marine Ecology Progress Series 475:155–166. Le Tourneux, F., and E. Bourget. 1988. Importance of physical and biological settlement cues

167 used at different spatial scales by the larvae of Semibalanus balanoides. Marine Biology 97:57–66. Lee, C. G., A. P. Farrell, A. Lotto, M. J. MacNutt, S. G. Hinch, and M. C. Healey. 2003. The effect of temperature on swimming performance and oxygen consumption in adult sockeye (Oncorhynchus nerka) and coho (O. kisutch) salmon stocks. Journal of Experimental Biology 206:3239–3251. Lemoine, N. P., W. A. Drews, D. E. Burkepile, and J. D. Parker. 2013. Increased temperature alters feeding behavior of a generalist herbivore. Oikos 122:1669–1678. Leonard, G. H. 2000. Latitudinal variation in species interactions: A test in the New England rocky intertidal zone. Ecology 81:1015–1031. Lima, F. P., P. A. Ribeiro, N. Queiroz, S. J. Hawkins, and A. M. Santos. 2007. Do distributional shifts of northern and southern species of algae match the warming pattern? Global Change Biology 13:2592–2604. Lopez-Urrutia, A., E. San Martin, R. Harris, and X. Irigoien. 2006. Scaling the metabolic balance of the oceans. Proceedings of the National Academy of Sciences 103:8739–8744. Lubchenco, J. 1983. Littorina and Fucus: effects of herbivores, substratum heterogeneity, and plant escapes during succession. Ecology 64:1116–1123. Lurgi, M., B. C. Lopez, and J. M. Montoya. 2012. Novel communities from climate change. Philosophical Transactions of the Royal Society B: Biological Sciences 367:2913–2922. MacArthur, R. H. 1972. Geographical Ecology. Princeton University Press, Princeton, NH. Mancinelli, G. 2012. On the trophic ecology of Gammaridea (Crustacea: Amphipoda) in coastal waters: A European-scale analysis of stable isotopes data. Estuarine, Coastal, and Shelf Science 114:130–139. McKee, D., D. Atkinson, S. Collings, J. Eaton, A. Gill, I. Harvey, K. Hatton, T. Heyes, D. Wilson, and B. Moss. 2003. Response of freshwater microcosm communities to nutrients, fish, and elevated temperature during winter and summer. Limnology and Oceanography 48:707–722. McMahon, R. F. 1990. Thermal tolerance, evaporative water loss, air-water oxygen consumption and zonation of intertidal prosobranchs: a new synthesis 193:241–260. Menge, B. A. 1976. Organization of the New England rocky intertidal community: role of predation, competition, and environmental heterogeneity. Ecological Monographs 46:355–

168 393. Menge, B. A. 2000. Recruitment vs. postrecruitment processes as determinants of barnacle population abundance. Ecological Monographs 70:265–288. Menge, B. A., T. C. Gouhier, T. Freidenburg, and J. Lubchenco. 2011. Linking long-term, large- scale climatic and environmental variability to patterns of marine invertebrate recruitment: Toward explaining “unexplained” variation. Journal of Experimental Marine Biology and Ecology 400:236–249. Menge, B., F. Chan, and J. Lubchenco. 2008. Response of a rocky intertidal ecosystem engineer and community dominant to climate change. Ecology Letters 11:151–162. Menzel, A., and P. Fabian. 1999. Growing season extended in Europe. Nature 397:659–659. Mieszkowska, N., and S. J. Hawkins. 2007. Long-term changes in the geographic distribution and population structures of Osilinus lineatus (Gastropoda: Trochidae) in Britain and Ireland. Journal of the Marine Biological Association of the United Kingdom 87:537–545. Mieszkowska, N., M. A. Kendall, S. J. Hawkins, R. Leaper, P. Williamson, N. J. Hardman- Mountford, and A. J. Southward. 2006. Changes in the range of some common rocky shore species in Britain – a response to climate change? Hydrobiologia 555:241–251. Miller, K. M., and T. H. Carefoot. 1989. The role of spatial and size refuges in the interaction between juvenile barnacles and grazing limpets. Journal of Experimental Marine Biology and Ecology 134:157–174. Miller, K. M., S. M. Blower, D. Hedgecock, and J. Roughgarden. 1989. Comparison of larval and adult stages of Chthamalus dalli and Chthamalus fissus (Cirripedia: Thoracica). Journal of Crustacean Biology 9:242–256. Miller, L. P., and B. Gaylord. 2007. Barriers to flow: The effects of experimental cage structures on water velocities in high-energy subtidal and intertidal environments. Journal of Experimental Marine Biology and Ecology 344:215–228. Moise, E. R. D., and H. A. L. Henry. 2012. Interactions of herbivore exclusion with warming and N addition in a grass-dominated temperate old field. Oecologia 169:1127–1136. Moore, P. J., R. C. Thompson, and S. J. Hawkins. 2011. Phenological changes in intertidal con- specific gastropods in response to climate warming. Global Change Biology 17:709–719. Moore, P., R. C. Thompson, and S. J. Hawkins. 2007a. Effects of grazer identity on the probability of escapes by a canopy-forming macroalga. Journal of Experimental Marine

169 Biology and Ecology 344:170–180. Moore, P., S J Hawkins, and R. C. Thompson. 2007b. Role of biological habitat amelioration in altering the relative responses of congeneric species to climate change. Marine Ecology Progress Series 334:11–19. Morelissen, B., and C. D. G. Harley. 2007. The effects of temperature on producers, consumers, and plant-herbivore interactions in an intertidal community. Journal of Experimental Marine Biology and Ecology 348:162–173. Muren, U., J. Berglund, K. Samuelsson, and A. Andersson. 2005. Potential effects of elevated sea-water temperature on pelagic food webs. Hydrobiologia 545:153–166. Nasrolahi, A., C. Pansch, M. Lenz, and M. Wahl. 2011. Being young in a changing world: how temperature and salinity changes interactively modify the performance of larval stages of the barnacle Amphibalanus improvisus. Marine Biology 159:331–340. Nicotri, M. E. 1977. Grazing effects of four marine intertidal herbivores on the microflora. Ecology 58:1020–1032. O'Connor, M. I. 2009. Warming strengthens an herbivore-plant interaction. Ecology 90:388–398. O'Connor, M. I., E. R. Selig, M. L. Pinsky, and F. Altermatt. 2011. Toward a conceptual synthesis for climate change responses. Global Ecology and Biogeography 21:693–703. Oksanen, L., S. D. Fretwell, J. Arruda, and P. Niemela. 1981. Exploitation ecosystems in gradients of primary productivity. American Naturalist 118:240–261. Orfanidis, S. 1993. Temperature responses and distribution of several Mediterranean macroalgae belonging to different distribution groups. Botanica Marina 36:359–370. Ostrander, G. K., K. M. Armstrong, E. T. Knobbe, D. Gerace, and E. P. Scully. 2000. Rapid transition in the structure of a coral reef community: the effects of coral bleaching and physical disturbance. Proceedings of the National Academy of Sciences 97:5297–5302. Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist 100:65–75. Paine, R. T. 1981. Barnacle ecology: is competition important? The forgotten roles of disturbance and predation. Paleobiology 7:553–560. Park, T. 1954. Experimental studies of interspecies competition II. Temperature, humidity, and competition in two species of Tribolium. Physiological Zoology 28:177–238. Parmesan, C. 2007. Influences of species, latitudes and methodologies on estimates of phenological response to global warming. Global Change Biology 13:1860–1872.

170 Parmesan, C., and G. Yohe. 2003. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421:37–42. Peery, M. Z., R. J. Gutiérrez, R. Kirby, O. E. LeDee, and W. LaHaye. 2011. Climate change and spotted owls: potentially contrasting responses in the Southwestern United States. Global Change Biology 18:865–880. Perry, A., P. Low, J. Ellis, and J. Reynolds. 2005. Climate change and distribution shifts in marine fishes. Science 308:1912–1915. Petchey, O., P. McPhearson, T. Casey, and P. Morin. 1999. Environmental warming alters food- web structure and ecosystem function. Nature 402:69–72. Pfeiffer-Hoyt, A. S., and M. A. McManus. 2005. Modeling the effects of environmental variability on Balanus glandula larval development. Journal of Plankton Research 27:1211– 1228. Philippart, C., H. van Aken, J. Beukema, O. Bos, G. Cadee, and R. Dekker. 2003. Climate- related changes in recruitment of the bivalve Macoma balthica. Limnology and Oceanography 48:2171–2185. Pimm, S. L. 2009. Climate disruption and biodiversity. Current Biology 19:R595–R601. Pincebourde, S., E. Sanford, and B. Helmuth. 2008. Body temperature during low tide alters the feeding performance of a top intertidal predator. Limnology and Oceanography 53:1562– 1573. Pinsky, M. L., B. Worm, M. J. Fogarty, J. L. Sarmiento, and S. A. Levin. 2013. Marine taxa track local climate velocities. Science 341:1239–1242. Platt, T., C. Fuentes-Yaco, and K. Frank. 2003. Spring algal bloom and larval fish survival. Nature 423:398–399. Poloczanska, E. S., C. J. Brown, W. J. Sydeman, W. Kiessling, D. S. Schoeman, P. J. Moore, K. Brander, J. F. Bruno, L. B. Buckley, M. T. Burrows, C. M. Duarte, B. S. Halpern, J. Holding, C. V. Kappel, M. I. O'Connor, J. M. Pandolfi, C. Parmesan, F. Schwing, S. A. Thompson, and A. J. Richardson. 2013. Global imprint of climate change on marine life. Nature Climate Change 3:1–7. Poloczanska, E. S., S. J. Hawkins, A. J. Southward, and M. T. Burrows. 2008. Modeling the response of populations of competing species to climate change. Ecology 89:3138–3149. Pomerat, C. M., and E. R. Reiner. 1942. The influence of surface angle and of light on the

171 attachment of barnacles and other sedentary organisms. Biological Bulletin 82:14–25. Poore, A. G. B., A. Graba-Landry, M. Favret, H. S. Brennand, M. Byrne, and S. A. Dworjanyn. 2013. Direct and indirect effects of ocean acidification and warming on a marine plant– herbivore interaction. Oecologia 173:1113–1124. Poore, A. G. B., A. H. Campbell, R. A. Coleman, G. J. Edgar, V. Jormalainen, P. L. Reynolds, E. E. Sotka, J. J. Stachowicz, R. B. Taylor, M. A. Vanderklift, and J. E. Duffy. 2012. Global patterns in the impact of marine herbivores on benthic primary producers. Ecology Letters 15:912–922. Post, E. 2013. Erosion of community diversity and stability by herbivore removal under warming. Proceedings of the Royal Society of London B: Biological Sciences 280:20122722–20122722. Post, E., and C. Pedersen. 2008. Opposing plant community responses to warming with and without herbivores. Proceedings of the National Academy of Sciences 105:12353–12358. Pounds, J., M. Fogden, and J. Campbell. 1999. Biological response to climate change on a tropical mountain. Nature 398:611–615. Pörtner, H. O. 2010. Oxygen- and capacity-limitation of thermal tolerance: a matrix for integrating climate-related stressor effects in marine ecosystems. Journal of Experimental Biology 213:881–893. Prieto, P., J. Peñuelas, F. Lloret, L. Llorens, and M. Estiarte. 2009. Experimental drought and warming decrease diversity and slow down post-fire succession in a Mediterranean shrubland. Ecography 32:623–636. Quinn, G. P., and M. J. Keough. 2002. Experimental design and data analysis for biologists. Cambridge University Press Cambridge, Cambridge, UK. Radchuk, V., C. Turlure, and N. Schtickzelle. 2013. Each life stage matters: the importance of assessing the response to climate change over the complete life cycle in butterflies. Journal of Animal Ecology 82:275–285. Richardson, A. J. 2008. In hot water: zooplankton and climate change. ICES Journal of Marine Science 65:279–295. Richardson, A. J., and E. S. Poloczanska. 2008. Under-resourced, under threat. Science 320:1294–1295. Riera, P., L. Stal, and J. Nieuwenhuize. 2004. Utilization of food sources by invertebrates in a

172 man-made intertidal ecosystem (Westerschelde, the Netherlands): a δ13C and δ15N study. Journal of the Marine Biological Association of the United Kingdom 84:323–326. Root, T., J. Price, K. Hall, S. Schneider, C. Rosenzweig, and J. Pounds. 2003. Fingerprints of global warming on wild animals and plants. Nature 421:57–60. Roy, K., D. Jablonsk, and J. W. Valentine. 1995. Thermally anomalous assemblages revisited: patterns in the extraprovincial latitudinal range shifts of Pleistocene marine mollusks. Geology 23:1071–1074. Roy, K., D. Jablonski, and J. Valentine. 2001. Climate change, species range limits and body size in marine bivalves. Ecology Letters 4:366–370. Rudolf, V. H. W., and M. Singh. 2013. Disentangling climate change effects on species interactions: effects of temperature, phenological shifts, and body size. Oecologia 173:1043– 1052. Ruesink, J. 2000. Intertidal mesograzers in field microcosms: linking laboratory feeding rates to community dynamics. Journal of Experimental Marine Biology and Ecology 248:163–176. Sagarin, R. D., J. P. Barry, S. Gilman, and C. H. Baxter. 1999. Climate-related change in an intertidal community over short and long time scales. Ecological Monographs 69:465–490. Sampath-Wiley, P., C. D. Neefus, and L. S. Jahnke. 2008. Seasonal effects of sun exposure and emersion on intertidal seaweed physiology: Fluctuations in antioxidant contents, photosynthetic pigments and photosynthetic efficiency in the red alga Porphyra umbilicalis Kützing (Rhodophyta, Bangiales). Journal of Experimental Marine Biology and Ecology 361:83–91. Sanford, E. 1999. Regulation of keystone predation by small changes in ocean temperature. Science 283:2095–2097. Sanford, E. 2002. Water temperature, predation, and the neglected role of physiological rate effects in rocky intertidal communities. Integrative and Comparative Biology 42:881–891. SAS Institute Inc. 2009. SAS/STAT 9.2 User's Guide. second edition. SAS Institute Inc, Cary, North Carolina. Satheesh, S., and S. G. Wesley. 2010. Influence of substratum colour on the recruitment of macrofouling communities. Journal of the Marine Biological Association of the United Kingdom 90:941–946. Schiel, D. R., J. R. Steinbeck, and M. S. Foster. 2004. Ten years of induced ocean warming

173 causes comprehensive changes in marine benthic communities. Ecology 85:1833–1839. Schroter, D. 2005. Ecosystem service supply and vulnerability to global change in Europe. Science 310:1333–1337. Sheridan, J. A., and D. Bickford. 2011. Shrinking body size as an ecological response to climate change. Nature Climate Change 1:401–406. Sih, A., G. Englund, and D. Wooster. 1998. Emergent impacts of multiple predators on prey. Trends in Ecology & Evolution 13:350–355. Sims, D. W., M. J. Genner, A. J. Southward, and S J Hawkins. 2001. Timing of squid migration reflects North Atlantic climate variability. Proceedings of the Royal Society of London B: Biological Sciences 268:2607–2611. Sims, D., V. Wearmouth, M. Genner, A. Southward, and S. Hawkins. 2004. Low‐temperature‐ driven early spawning migration of a temperate marine fish. Journal of Animal Ecology 73:333–341. Smale, D. A., and T. Wernberg. 2012. Short-term in situ warming influences early development of sessile assemblages. Marine Ecology Progress Series 453:129–136. Smale, D. A., T. Wernberg, L. S. Peck, and D. K. A. Barnes. 2011. Turning on the heat: Ecological response to simulated warming in the sea. PLoS ONE 6:e16050. Smith, C. M., and J. A. Berry. 1986. Recovery of photosynthesis after exposure of intertidal algae to osmotic and temperature stresses: comparative studies of species with differing distributional limits. Oecologia 70:6–12. Somero, G. N. 2002. Thermal physiology and vertical zonation of intertidal animals: optima, limits, and costs of living. Integrative and Comparative Biology 42:780–789. Somero, G. N. 2007. Heat Stress. Pages 266–270 in M. W. Denny and S. D. Gaines, editors. Encyclopedia of Tidepools and Rocky Shores. University of California Press, Los Angeles. Sorte, C. J. B., S. L. Williams, and J. T. Carlton. 2010. Marine range shifts and species introductions: comparative spread rates and community impacts. Global Ecology and Biogeography 19:303–316. Sotka, E. E., J. P. Wares, J. A. Barth, R. K. Grosberg, and S. R. Palumbi. 2004. Strong genetic clines and geographical variation in gene flow in the rocky intertidal barnacle Balanus glandula. Molecular Ecology 13:2143–2156. Southward, A. J. 1991. Forty years of changes in species composition and population density of

174 barnacles on a rocky shore near Plymouth. Journal of the Marine Biological Association of the United Kingdom 71:495–513. Southward, A. J., and D. J. Crisp. 1954. Recent changes in the distribution of the intertidal barnacles Chthamalus stellatus Poli and Balanus balanoides L. in the British Isles. The Journal of Animal Ecology 23:163–177. Southward, A. J., and E. C. Southward. 1967. On the biology of an intertidal chthamalid (Crustacea, Cirripedia) from the Chukchi Sea. Arctic:8–20. Southward, A. J., O. Langmead, N. J. Hardman-Mountford, J. Aiken, G. T. Boalch, P. R. Dando, M. J. Genner, I. Joint, M. A. Kendall, N. C. Halliday, R. P. Harris, R. Leaper, N. Mieszkowska, R. D. Pingree, A. J. Richardson, D. W. Sims, T. Smith, A. W. Walne, and S. J. Hawkins. 2005. Long-term oceanographic and ecological research in the Western English Channel. Advances in Marine Biology 47:1–105. Southward, A. J., S. J. Hawkins, and M. T. Burrows. 1995. Seventy years' observations of changes in distribution and abundance of zooplankton and intertidal organisms in the western English Channel in relation to rising sea temperature. Journal of Thermal Biology 20:127–155. Stachowicz, J., J. Terwin, R. Whitlatch, and R. Osman. 2002. Linking climate change and biological invasions: Ocean warming facilitates nonindigenous species invasions. Proceedings of the National Academy of Sciences 99:15497–15500. Steffensen, D. A. 1976. The effect of nutrient enrichment and temperature on the growth in culture of Ulva lactuca L. Aquatic Botany 2:337–351. Steinbeck, J. R., D. R. Schiel, and M. S. Foster. 2005. Detecting long-term change in complex communities: a case study from the rocky intertidal zone. Ecological Applications 15:1813– 1832. Steinke, S., P. S. Yu, M. Kucera, and M. T. Chen. 2008. No-analog planktonic foraminiferal faunas in the glacial southern South China Sea: Implications for the magnitude of glacial cooling in the western Pacific warm pool. Marine Micropaleontology 66:71–90. Steneck, R. S., and L. Watling. 1982. Feeding capabilities and limitation of herbivorous molluscs: a functional group approach. Marine Biology 68:299–319. Steneck, R. S., S. D. Hacker, and M. N. Dethier. 1991. Mechanisms of competitive dominance between crustose coralline algae: an herbivore-mediated competitive reversal. Ecology:938–

175 950. Stewart, R. I., M. Dossena, D. A. Bohan, E. Jeppesen, R. L. Kordas, M. E. Ledger, M. Meerhoff, B. Moss, C. Mulder, and J. B. Shurin. 2013. Mesocosm experiments as a tool for ecological climate-change research. Advances in Ecological Research 48:71–181. Stickle, W. B., and G. J. Denoux. 1976. Effects of in situ tidal salinity fluctuations on osmotic and lonic composition of body fluid in Southeastern Alaska rocky intertidal fauna. Marine Biology 37:125–135. Stillman, J. H., and G. N. Somero. 2000. A comparative analysis of the upper thermal tolerance limits of Eastern Pacific porcelain crabs, genus Petrolisthes: influences of latitude, vertical zonation, acclimation, and phylogeny. Physiological and Biochemical Zoology 73:200–208. Sunday, J. M., A. E. Bates, M. R. Kearney, R. K. Colwell, N. K. Dulvy, J. T. Longino, and R. B. Huey. 2014. Thermal-safety margins and the necessity of thermoregulatory behavior across latitude and elevation. Proceedings of the National Academy of Sciences 111:5610–5615. Sutton, R. T., B. Dong, and J. M. Gregory. 2007. Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. Geophysical Research Letters 34. Swain, G., S. Herpe, E. Ralston, and M. Tribou. 2006. Short-term testing of antifouling surfaces: the importance of colour. Biofouling 22:425–429. Tamburello, L., F. Bulleri, I. Bertocci, E. Maggi, and L. Benedetti-Cecchi. 2013. Reddened seascapes: experimentally induced shifts in 1/f spectra of spatial variability in rocky assemblages. Ecology 94:1102–1111. Teoh, M.-L., S.-M. Phang, and W.-L. Chu. 2012. Response of Antarctic, temperate, and tropical microalgae to temperature stress. Journal of Applied Phycology 25:285–297. Thomas, C. D., A. Cameron, R. E. Green, M. Bakkenes, L. J. Beaumont, Y. C. Collingham, B. F. N. Erasmus, M. F. De Siqueira, A. Grainger, L. Hannah, L. Hughes, B. Huntley, A. S. Van Jaarsveld, G. F. Midgley, L. Miles, M. A. Ortega-Huerta, A. T. Peterson, O. L. Phillips, and S. E. Williams. 2004. Extinction risk from climate change. Nature 427:145–148. Thomas, C., B. Crear, and P. Hart. 2000. The effect of temperature on survival, growth, feeding and metabolic activity of the southern rock lobster, Jasus edwardsii. Aquaculture 185:73–84. Thompson, R., and R. Clark. 2008. Is spring starting earlier? The Holocene 18:95–104. Thompson, R., T. Norton, and S. Hawkins. 2004. Physical stress and biological control regulate

176 the producer-consumer balance in intertidal biofilms. Ecology 85:1372–1382. Tomanek, L., and E. Sanford. 2003. Heat-shock protein 70 (Hsp70) as a biochemical stress indicator: an experimental field test in two congeneric intertidal gastropods (Genus: Tegula). Biological Bulletin 205:276–284. Trenberth, K. E. 2012. Framing the way to relate climate extremes to climate change. Climatic Change 115:283–290. Underwood, A. J. 1980. The effects of grazing by gastropods and physical factors on the upper limits of distribution of intertidal macroalgae. Oecologia 46:201–213. van Tamelen, P., and M. S. Stekoll. 1997. The role of barnacles in the recruitment and subsequent survival of the brown alga, Fucus gardneri (Silva). Journal of Experimental Marine Biology and Ecology 208:227–238. Vasseur, D. A., J. P. DeLong, B. Gilbert, H. S. Greig, C. D. G. Harley, K. S. McCann, V. Savage, T. D. Tunney, and M. I. O'Connor. 2014. Increased temperature variation poses a greater risk to species than climate warming. Proceedings of the Royal Society of London B: Biological Sciences 281:20132612–20132612. Visser, M. E., and C. Both. 2005. Review. Shifts in phenology due to global climate change: the need for a yardstick. Proceedings of the Royal Society of London B: Biological Sciences 272:2561–2569. VonEnde, C. N. 2001. Repeated-measures analysis. in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Second edition. Oxford University Press. Walther, G. R. 2002. Ecological responses to recent climate change. Nature 416:389–395. Wang, S., J. Duan, G. Xu, Y. Wang, Z. Zhang, Y. Rui, C. Luo, B. Xu, X. Zhu, X. Chang, X. Cui, H. Niu, X. Zhao, and W. Wang. 2012. Effects of warming and grazing on soil N availability, species composition, and ANPP in an alpine meadow. Ecology 93:2365–2376. Watanuki, Y., M. Ito, T. Deguchi, and S. Minobe. 2009. Climate-forced seasonal mismatch between the hatching of rhinoceros auklets and the availability of anchovy. Marine Ecology Progress Series 393:259–271. Webster, N. S., R. Soo, R. Cobb, and A. P. Negri. 2010. Elevated seawater temperature causes a microbial shift on crustose coralline algae with implications for the recruitment of coral larvae. The ISME Journal 5:759–770. Wells, B. W. 1928. Plant communities of the coastal plain of North Carolina and their

177 successional relations. Ecology 9:230–242. Wernberg, T., D. A. Smale, and M. S. Thomsen. 2012a. A decade of climate change experiments on marine organisms: procedures, patterns and problems. Global Change Biology 18:1491– 1498. Wernberg, T., D. A. Smale, F. Tuya, M. S. Thomsen, T. J. Langlois, T. de Bettignies, S. Bennett, and C. S. Rousseaux. 2012b. An extreme climatic event alters marine ecosystem structure in a global biodiversity hotspot. Nature Climate Change 3:78–82. Wernberg, T., M. S. Thomsen, and F. Tuya. 2011. Biogenic habitat structure of seaweeds change along a latitudinal gradient in ocean temperature. Journal of Experimental Marine Biology and Ecology 400:264–271. Wernberg, T., M. S. Thomsen, F. Tuya, G. A. Kendrick, P. A. Staehr, and B. D. Toohey. 2010. Decreasing resilience of kelp beds along a latitudinal temperature gradient: potential implications for a warmer future. Ecology Letters 13:685–694. Werner, F. J., and B. Matthiessen. 2012. Temperature indirectly affects benthic microalgal diversity by altering effects of top-down but not bottom-up control. Oikos 122:52–63. Wethey, D. 1983. Geographic limits and local zonation: the barnacles Semibalanus (Balanus) and Chthamalus in New England. Biological Bulletin 165:330–341. Wethey, D. 1984a. Sun and shade mediate competition in the barnacles Chthamalus and Semibalanus: a field experiment. Biological Bulletin 167:176–185. Wethey, D. S. 1984b. Effects of crowding on fecundity in barnacles: Semibalanus (Balanus) balanoides, Balanus glandula, and Chthamalus dalli. Canadian Journal of Zoology 62:1788– 1795. Wethey, D. S., and S. A. Woodin. 2008. Ecological hindcasting of biogeographic responses to climate change in the European intertidal zone. Hydrobiologia 606:139–151. Williams, G. A. 1994. The relationship between shade and molluscan grazing in structuring communities on a moderately-exposed tropical rocky shore. Journal of Experimental Marine Biology and Ecology 178:79–95. Williams, J. W., and S. T. Jackson. 2007. Novel climates, no-analog communities, and ecological surprises. Frontiers in Ecology and the Environment 5:475–482. Williams, S. L., M. E. S. Bracken, and E. Jones. 2013. Additive effects of physical stress and herbivores on intertidal seaweed biodiversity. Ecology 94:1089–1101.

178 Wilson, B. 2006. British Columbia Coast and Marine Environment Project 2006:1–34. Wolcott, T. G. 1973. Physiological ecology and intertidal zonation in limpets (Acmaea): a critical look at “limiting factors.” Biological Bulletin 145:389–422. Wolkovich, E. M., B. I. Cook, J. M. Allen, T. M. Crimmins, J. L. Betancourt, S. E. Travers, S. Pau, J. Regetz, T. J. Davies, N. J. B. Kraft, T. R. Ault, K. Bolmgren, S. J. Mazer, G. J. McCabe, B. J. McGill, C. Parmesan, N. Salamin, M. D. Schwartz, and E. E. Cleland. 2012. Warming experiments underpredict plant phenological responses to climate change. Nature:1–4. Wonham, M., M. I. O'Connor, and C. D. G. Harley. 2005. Positive effects of a dominant invader on introduced and native mudflat species. Marine Ecology Progress Series 289:109–116. Wood, S. A., S. A. Lilley, D. R. Schiel, and J. B. Shurin. 2010. Organismal traits are more important than environment for species interactions in the intertidal zone. Ecology Letters 13:1160–1171. Woodward, G., J. B. Dybkjaer, J. S. Olafsson, G. M. Gislason, E. R. Hannesdóttir, and N. Friberg. 2010. Sentinel systems on the razor's edge: effects of warming on Arctic geothermal stream ecosystems. Global Change Biology 16:1979–1991. Yamane, L., and S. Gilman. 2009. Opposite responses by an intertidal predator to increasing aquatic and aerial temperatures. Marine Ecology Progress Series 393:27–36.

179 Appendix A Thermal Performance of Salt Spring Species

Algal Species Morphology Response Type Temp Medium Population Intertidal Distribution Reference (°C) Zone Acrosiphonia spp. prostrate growth optimum 15 water Friday Harbor, low-mid Oregon – Bering Sea (Bischoff and Wiencke (A. arcta) branches growth critical 20 water WA 1995b) survival lethal 22-23 water Analipus japonicus semi-upright nd low-mid Pt. Conception, CA – branches Aleutian Islands, AK Colpomenia peregrina upright, growth optimum 10-25 n/g Mediterranean low S. California – Aleutian (Orfanidis 1993) globular Sea Islands, AK Fucus distichus semi-upright nd low-high Cen. California – Arctic blades Ocean Halosaccion upright sac- nd low-mid Pt. Conception, CA – glandiforme like Aleutian Islands, AK Hildenbrandia rubra crust photo. reduced 27 water Nova Scotia low-high Panama – Aleutian Islands, (Kim and Garbary AK 2006) Mastocarpus upright blades / photo. critical 35 air Hopkins, CA mid-high N. California – Aleutian (Bell 1993) papillatus crust Islands, AK Mazzaella splendens prostrate sheets nd low Baja California – Southeast AK Microcladia borealis prostrate nd low-mid San Luis Obispo, CA – filaments Aleutian Islands, AK Navicula spp. prostrate cells growth optimum 11 water culture: San low-mid worldwide (Teoh et al. 2012) (N. incerta) growth critical 33 water Francisco, CA Other colonial benthic prostrate low-high many spp. worldwide diatoms Petalonia fascia prostrate growth optimum 16 water n/g low-mid Baja California – Arctic (Hsiao 1970) blades growth critical 26-31 Ocean Polysiphonia spp. prostrate nd low Baja California – Aleutian (P. pacifica) filaments Islands, AK Pyropia spp. prostrate sheets photo. optimum 29 water Hopkins, CA mid Baja California – Kodiak (Smith and Berry 1986) (P. perforata) photo. lethal 33 water Island, AK Ulothrix spp. prostrate nd mid-high S. California – Arctic (U. flacca) filaments Ocean Ulva spp. prostrate sheets growth optimum 16 water Christchurch, low-high worldwide (Steffensen 1976) (U. lactuca) growth critical 25 water New Zealand Urospora spp prostrate growth optimum 5 water San Juan, WA mid-high Cen. California – S. Alaska (Bischoff and Wiencke (U. penicilliformis) filaments growth critical 20 water 1995a) survival lethal 24 water Table A.1: Salt Spring algal species thermal information

180 Invertebrate Species Response Type Temp Medium Population Zone NE Pacific Reference (°C) Distribution Amphipods nd for NE Pacific spp low-high many spp. worldwide

Balanus glandula irreversible protein critical >34 air Charleston, OR high N. Baja – S. Alaska (Berger and Emlet 2007) damage survival lethal 44 air Vancouver, BC (Liao & Harley unpub) Chthamalus dalli cirri beating optimum 28 water Puget Sound, WA high S. CA - Alaska (Southward and Southward cirri beating lethal 35 water 1967) survival lethal 46 air Vancouver, BC (Liao & Harley unpub) Emplectonema gracile nd low-mid Chile – Aleutian Islands, AK Littorina plena nd high Baja – S. Alaska

Littorina scutulata nd high Baja – S. Alaska

Lottia digitalis survival lethal 38-39 water Bodega Bay, CA high Pt. Conception, CA – S. (Wolcott 1973) survival lethal 43-44 air Alaska Lottia paradigitalis nd high Pt. Conception, CA – S. Alaska Lottia pelta survival lethal 34-35 water Bodega Bay, CA low-mid Baja – S. Alaska (Wolcott 1973) survival lethal 39-40 air Lottia scutum survival lethal 35-36 water Bodega Bay, CA low-mid Pt. Conception, CA – (Wolcott 1973) survival lethal 40-41 air Bering Sea Nereid polychaete low-mid

Table A.2: Salt Spring invertebrate species thermal information. Species name: species identified in Salt Spring plots (in parentheses: related species for which thermal data were available) Response types: Optimum = temperature where a functional trait is maximized. Critical (sublethal) = mean temperature at which individuals lose some essential function (e.g. growth). Lethal = temperatures where a predefined percentage of individuals die after a fixed duration of exposure (e.g. LT50) n/g = not given in paper, nd = no data for this species (or similar species from the same geographic region) Table references can be found in the main bibliography.

181 Appendix B Linear Mixed Effect Model Results

Model AICc Δι AIC ωi ωi/ωj Model AICc Δι AIC ωi ωi/ωj (a) Total Algae percent cover (b) Green algae percent cover Transformation, inverse 10th root: Y' =(( Y-0.1)-1)/(-0.1) Transformation, inverse 5th root: Y' =(( Y-0.2)-1)/(-0.2) Temp, zone, date, all 2-ways, 3-way 835.7 0 1 60.34 Temp, zone, date, all 2-ways, 3-way 520.6 0 0.99 8.13 x 106 Temp, zone, date, all 2-ways 885.7 50.0 1.39 x 10-11 Zone, date, Z x D 553.4 32.8 7.54 x 10-8 Zone, date, Z x D 887.0 51.3 7.25 x 10-12 Temp, zone, date, Z x D 555.3 34.7 2.92 x 10-8 Temp, zone, date, Z x D 890.3 54.6 1.39 x 10-12 Temp, zone, date, T x Z, Z x D 556.7 36.1 1.45 x 10-8 Temp, zone, date, T x Z, Z x D 892.9 57.2 3.79 x 10-13 Temp, zone, date, all 2-ways 559.3 38.7 3.95 x 10-9 Temp, zone, date, T x D 1014.9 179.2 1.22 x 10-39 Temp, zone, date, T x D 665.3 144.7 3.79 x 10-32 Temp, zone, date, T x Z, T x D 1018.1 182.4 2.47 x 10-40 Temp, zone, date, T x Z, T x D 666.5 145.9 2.08 x 10-32 Zone, date 1022.9 187.2 2.24 x 10-41 Zone, date 668.2 147.6 8.89 x 10-33 Temp, zone, date 1026.6 190.9 3.52 x 10-42 Temp, zone, date 672.5 151.9 1.04 x 10-33 Temp, zone, date, T x Z 1029.3 193.6 9.13 x 10-43 Temp, zone, date, T x Z 672.6 152.0 9.85 x 10-34 Temp, date, T x D 1071.1 235.4 7.65 x 10-52 Date 693.5 172.9 2.85 x 10-38 Date 1081.0 245.3 5.42 x 10-54 Temp, date 696.1 175.5 7.77 x 10-39 Temp, date 1083.0 247.3 1.99 x 10-54 replicate plate 781.7 261.1 2.01 x 10-57 replicate plate 1142.2 306.5 2.78 x 10-67 Zone 785.9 265.3 2.46 x 10-58 Zone 1181.0 345.3 1.04 x 10-75 Temp, zone 789.1 268.5 4.97 x 10-59 Temp, zone, T x Z 1182.4 346.7 5.19 x 10-76 Temp, zone, T x Z 789.5 268.9 4.07 x 10-59 Temp, zone 1182.6 346.9 4.69 x 10-76 Temp 883.2 362.6 1.83 x 10-79 Temp 1272.3 436.6 1.56 x 10-95 Temp, date, T x D not estimable

Table B.1: Results from linear mixed effect model analyses. For each functional group response, all 18 possible models are ranked by the lowest Corrected Akaike information criterion (AICc) score. Akaike weights (wi) were calculated from the differences between AICc’s (Di) for each model and the model with the minimum AICc. The relative likelihood of the best-fitting model to the next best, or any other model(s) was calculated from the evidence ratio of respective model weights (wi/wj). The transformation used for each functional group is given in the first line under the group name. Transformations were determined using the Box-Cox transformation in SAS.

182

Model AICc Δι AIC ωi ωi/ωj Model AICc Δι AIC ωi ωi/ωj (c) Grazer density (d) Barnacle density Transformation, inverse square: Y' =(( Y-2)-1)/(-2) Transformation, 5th root: Y' =(( Y0.2)-1)/(0.2) Temp -1167.5 0 0.97 36.00 Temp, zone, date, all 2-ways, 3-way 1304.4 0 1 1.80 x 1023 Zone -1160.3 7.2 0.03 Temp, zone, date, all 2-ways 1411.5 107.1 5.54 x 10-24 Temp, zone -1152.1 15.4 4.41 x 10-4 Temp, zone, date, T x Z, Z x D 1451.2 146.8 1.33 x 10-32 Temp, zone, T x Z -1138.1 29.4 4.02 x 10-7 Zone, date, Z x D 1468.4 164.0 2.44 x 10-36 Date -1058.1 109.4 1.71 x 10-24 Temp, zone, date, Z x D 1469.5 165.1 1.41 x 10-36 Temp, date -1048.8 118.7 1.63 x 10-26 replicate plate 1472.6 168.2 2.99 x 10-37 Zone, date -1042.3 125.2 6.33 x 10-28 Temp, zone, date, T x Z, T x D 1500.9 196.5 2.14 x 10-43 Temp, zone, date -1033.5 134.0 7.77 x 10-30 Temp, zone, date, T x D 1512.3 207.9 7.16 x 10-46 Temp, zone, date, T x Z -1018.3 149.2 3.89 x 10-33 Temp, date, T x D 1516.7 212.3 7.94 x 10-47 Temp, date, T x D -1009.5 158.0 4.77 x 10-35 Temp, zone, date, T x Z 1534.0 229.6 1.39 x 10-50 Temp, zone, date, T x D -995.9 171.6 5.32 x 10-38 Zone, date 1549.2 244.8 6.96 x 10-54 Temp, zone, date, T x Z, T x D -980.6 186.9 2.53 x 10-41 Temp, zone, date 1549.6 245.2 5.69 x 10-54 Zone, date, Z x D -949.7 217.8 4.94 x 10-48 Temp, date 1553.8 249.4 6.97 x 10-55 replicate plate -949.5 218.0 4.47 x 10-48 Date 1559.5 255.1 4.03 x 10-56 Temp, zone, date, Z x D -941.0 226.5 6.37 x 10-50 Temp, zone, T x Z 1753.7 449.3 2.73 x 10-98 Temp, zone, date, T x Z, Z x D -925.3 242.2 2.48x 10-53 Zone 1763.6 459.2 1.93 x 10-100 Temp, zone, date, all 2-ways -895.7 271.8 9.28 x 10-60 Temp, zone 1763.6 459.2 1.93 x 10-100 -68 Temp, zone, date, all 2-ways, 3-way -856.5 311.0 2.85 x 10 1767.5 463.1 2.75 x 10-101

Table B.1: continued

183

Model AICc Δι AIC ωi ωi/ωj (e) Species Richness Untransformed Temp, zone, date, all 2-ways, 3-way 962.0 0 1 2.8 x 109 Temp, zone, date, all 2-ways 1005.5 43.5 3.58 x 10-10 Temp, zone, date, Z x D 1018.0 56.0 6.91 x 10-13 Temp, zone, date, T x Z, Z x D 1018.9 56.9 4.41 x 10-13 Zone, date, Z x D 1025.2 63.2 1.89 x 10-14 Temp, zone, date, T x Z, T x D 1080.5 118.5 1.85 x 10-26 Temp, zone, date, T x D 1081.1 119.1 1.37 x 10-26 Temp, date 1083.0 121.0 5.31 x 10-27 Temp, zone, date, T x Z 1094.5 132.5 1.69 x 10-29 Temp, zone, date 1094.9 132.9 1.38 x 10-29 Zone, date 1099.3 137.3 1.53 x 10-30 Temp, date, T x D 1103.9 141.9 1.54 x 10-31 Temp, zone 1116.9 154.9 2.31 x 10-34 Date 1119.8 157.8 5.42 x 10-35 replicate plate 1142.2 180.2 7.41 x 10-40 Zone 1181.0 219.0 2.78 x 10-48 Temp, zone, T x Z 1182.4 220.4 1.38 x 10-48 Temp 1272.3 310.3 4.16 x 10-68

Table B.1: continued

184 Appendix C Summary of Barnacle Thermal Responses

Parameter Spp Zone Trmt Mean lower CI upper CI n Effect of warming Density B. glandula High Warm 4.56 2.25 6.87 68 - 82.33% (back transformed B. glandula High Cool 25.83 23.55 28.12 58 from log(x+1) C. dalli High Warm 11.55 9.19 13.91 68 - 62.14% in RM-ANOVA) C. dalli High Cool 32.96 30.53 35.39 58

B. glandula Mid Warm 7.71 5.31 10.11 61 - 92.27% B. glandula Mid Cool 99.81 97.60 102.00 44 C. dalli Mid Warm 6.86 4.38 9.36 61 - 79.93% C. dalli Mid Cool 34.20 31.80 36.60 44

Percent Cover B. glandula High Warm 0.42 0.22 0.62 68 - 88.71% (back transformed B. glandula High Cool 3.72 2.407 5.27 66 from log(x+1) C. dalli High Warm 0.92 0.57 1.31 68 - 71.69% in RM-ANOVA) C. dalli High Cool 3.25 2.09 4.58 66

B. glandula Mid Warm 0.86 0.51 1.25 62 - 94.31% B. glandula Mid Cool 15.11 10.86 20.01 53 C. dalli Mid Warm 0.89 0.50 1.34 62 - 66.79% C. dalli Mid Cool 2.68 1.74 3.76 53

Survivorship B. glandula High Warm 0 0 0 2 - 100.00% June-Sept B. glandula High Cool 48.04 34.38 61.78 7 (back transformed C. dalli High Warm 41.41 6.29 79.00 5 - 58.13% from arcsine sqrt C. dalli High Cool 98.91 92.58 101.05 6 in RM-ANOVA) B. glandula Mid Warm 0.34 -0.33 2.34 6 - 87.17% B. glandula Mid Cool 2.65 -2.54 17.70 6 C. dalli Mid Warm 3.67 -2.45 14.84 6 + 54.85% C. dalli Mid Cool 2.37 -2.28 15.85 6

Growth B. glandula High Warm 0.111 0.031 0.193 2 - 16.5% (ns) Summer B. glandula High Cool 0.133 0.076 0.190 7 (back transformed C. dalli High Warm 0.027 0.021 0.033 7 - 50.91% from log(x+1) C. dalli High Cool 0.055 0.039 0.071 7 in ANOVA) B. glandula Mid Warm 0.103 0.030 0.177 4 - 42.5% (ns) B. glandula Mid Cool 0.179 0.140 0.220 6 C. dalli Mid Warm 0.043 0.019 0.067 6 - 33.85% C. dalli Mid Cool 0.065 0.047 0.083 6

Table C.1: Parameters of barnacle populations across one year (April 2009 – April 2010). Density and percent cover were highly variable over time (see Figures 4.3 and 4.4). Summary statistics and percentages are solely meant for ease of comparison. Confidence intervals (CI) encompass 95% of the data. The sample size (n) for density and percent cover represents the number of plates per treatment (initially = 7) x the number of sampling dates (10). Values less than 70 are due to loss of replicates over time. For survivorship and growth rates, n = the number of replicate plates. Statistics are given as in graphs (back transformed where appropriate). Percentage is listed when differences were statistically significant in ANOVA or RM-ANOVA, (ns) = not significant. Bolded percentages represent the difference between cool and warm treatments when CI are not overlapping.

185 Appendix D Taxonomic Responses to Warming and Limpets

Figure D.1: Abundance (back transformed) of key taxa through time (+/- SE). Density of invertebrates on left, percent cover of algal groups on right. Note the different y-axis scales. The estimates of total algae (h) included some rare species not shown individually in graphs (see Table F.1 for complete species list) and exceeds 100% cover because estimates included layers of algae. There are several dates when only a subset of taxa were sampled. For example, algae percent cover was quantified on all surveys, but invertebrates were not surveyed in September because photographs were used, which tend to underestimate invertebrate counts since they remain hidden in or under algae. Note differences in y-axis scales.

186

Figure D.1: Continued

187 Appendix E Species Contributing to Differences Between Treatments

Date / Treatment Species Av. Abund Av. Abund Av. Sim (Dis)Sim / SD Contr. (%) Cum. (%) (a) Summer 2011 (July 27, 2011) warm cool Cool Diatom film 3.79 38.96 4.80 49.88 49.88 sim = 78.10 B. glandula 2.39 24.48 3.17 31.34 81.22 Ulva sp. 1.85 13.99 1.11 17.91 99.13

Warm Diatom film 4.04 54.94 1.93 83.18 83.18 sim = 66.05 B. glandula 0.74 4.63 0.51 7.01 90.20

Cool vs. Warm Ulva sp. 0.22 1.85 12.57 1.54 29.56 29.56 disssim = 42.51 B. glandula 0.74 2.39 11.87 1.74 27.91 57.48 Diatom film 4.04 3.79 7.54 0.91 17.73 75.20 Gr. fil. algae 0.75 0.20 5.15 0.56 12.12 87.33 Pyropia sp. 0.49 0.00 2.65 0.45 6.24 93.57

excl con Control Diatom film 3.86 45.71 3.41 65.64 65.64 sim = 69.63 B. glandula 1.86 18.77 1.41 26.95 92.59

Exclusion Diatom film 3.97 48.19 1.68 64.67 64.67 sim = 74.51 Ulva sp. 1.25 11.83 0.91 15.87 80.54 B. glandula 1.27 10.34 0.90 13.88 94.42

Con vs. Excl Diatom film 3.97 3.86 7.39 0.75 21.78 21.78 disssim = 33.95 Gr. fil. algae 0.95 0.00 6.76 0.60 19.92 41.70 B. glandula 1.27 1.86 6.64 1.05 19.57 61.27 Ulva sp. 1.25 0.82 6.60 0.94 19.43 80.70 C. dalli 0.04 0.56 3.44 0.81 10.15 90.85

Table E.1: Percentage contributions of individual species to observed similarity for each treatment, estimated using SIMPER analyses (when significant differences were detected using PERMANOVA, see Tables 5.6-5.8). The cumulative 90% of contributors to dissimilarities are shown. (Dis)Sim / SD: the average contribution divided by the standard deviation of those contributions across all pairs of samples making up the average. Cont: contribution of each species to differences between treatments; Cum: running total of the contribution to the observed dissimilarity. Numbers beneath treatment names are the total similarity for that treatment or dissimilarity for the treatment pair.

188 Date / Treatment Species Av. Abund Av. Abund Av. Sim (Dis)Sim / SD Contr. (%) Cum. (%) (b) Winter 2012 (April 9, 2012) excl con Control B. glandula 1.66 9.65 0.83 28.12 28.12 sim = 34.30 Ulva sp. 2.11 8.53 0.83 24.88 53.00 C. dalli 0.94 4.76 1.06 13.89 66.89

C. peregrina 1.18 3.93 0.69 11.47 78.35 Diatom film 1.41 3.78 0.42 11.03 89.38 Gr. fil. algae 0.93 1.48 0.23 4.30 93.69

Exclusion Ulva sp. 4.21 37.28 3.13 70.25 70.25 sim = 53.06 Navicula 1.54 5.33 0.51 10.04 80.29 Diatom film 1.36 3.14 0.37 5.92 86.21 P. fascia 0.75 2.86 0.45 5.40 91.60

Con vs. Excl Ulva sp. 4.21 2.11 11.87 1.08 17.96 17.96 dissim = 66.09 Diatom film 1.36 1.41 8.34 1.01 12.61 30.57 Navicula 1.54 0.38 7.49 0.86 11.34 41.91 B. glandula 0.87 1.66 6.30 1.13 9.54 51.45 C. peregrina 0.50 1.18 5.18 1.11 7.84 59.29 P. fascia 0.75 0.46 4.55 0.79 6.89 66.17 Gr. fil. algae 0.23 0.93 4.45 0.58 6.74 72.91 C. dalli 0.34 0.94 4.04 1.26 6.11 79.03 ‘Petrocelis’ 0.16 0.58 3.62 0.49 5.47 84.50 F. distichus 0.34 0.38 2.50 0.59 3.78 88.28 H. rubra 0.12 0.49 2.43 0.57 3.68 91.97

Table E.1: continued

189 Date / Treatment Species Av. Abund Av. Abund Av. Sim (Dis)Sim / SD Contr. (%) Cum. (%) (c) Summer 2012 (August 28, 2012) warm cool Cool Ulva sp. 3.32 26.76 1.36 45.55 45.55 sim = 58.78 B. glandula 2.86 25.49 2.17 43.38 88.93 C. dalli 0.87 3.65 0.47 6.21 95.14

Warm Diatom film 3.27 28.72 2.00 40.03 40.03 sim = 70.62 Ulva sp. 2.72 21.67 1.70 30.68 70.72 B. glandula 1.60 11.54 0.97 16.34 87.06 Ulva sp. 1.11 7.54 0.86 10.67 97.73

Cool vs. Warm Diatom film 3.27 0.71 16.19 2.02 32.14 32.14 disssim = 50.35 B. glandula 1.60 2.86 9.13 1.52 18.14 50.29 Ulva sp. 2.72 3.32 8.47 1.15 16.82 67.10 C. dalli 1.11 0.87 5.16 0.96 10.24 77.35 Gr. fil. algae 0.41 0.12 2.71 0.55 5.37 82.72 F. distichus 0.09 0.37 2.27 0.45 4.52 87.24 Amphipods 0.12 0.21 1.09 0.96 2.17 89.40 ‘Petrocelis’ 0.00 0.18 0.99 0.37 1.96 91.37

excl con Control B. glandula 2.57 23.07 1.73 39.03 39.03 sim = 74.51 Ulva sp. 2.20 14.50 1.21 24.54 63.58 C. dalli 1.71 10.75 1.14 18.18 81.76 Diatom film 1.93 10.38 0.83 17.57 99.33

Exclusion Ulva sp. 3.84 33.93 2.25 48.27 48.27 sim = 69.63 Diatom film 2.06 18.95 0.94 26.95 75.22 B. glandula 1.89 13.97 1.10 19.87 95.09

Con vs. Excl Ulva sp. 3.84 2.20 11.63 1.14 24.91 24.91 disssim = 33.95 C. dalli 0.27 1.71 8.16 1.54 17.48 42.39 B. glandula 1.89 2.57 7.79 1.35 16.69 59.08 Diatom film 2.06 1.93 7.32 0.99 15.67 74.76 Gr. fil. algae 0.41 0.12 2.77 0.54 5.92 80.68 F. distichus 0.34 0.12 2.31 0.43 4.95 85.63 Amphipods 0.24 0.09 1.24 1.00 2.66 88.29 Brown crust 0.00 0.18 0.92 0.37 1.96 90.25

Table E.1: continued

190 Appendix F Species List for Salt Spring Island

Taxonomic group Genus / Species Phaeophyta Acrosiphonia sp. Analipus japonicus Colpomenia peregrina Fucus distichus Halosaccion glandiforme Petalonia fascia Scytosiphon sp.

Rhodophyta Cryptopleura sp. Hildenbrandia rubra Mastocarpus papillatus crust (‘Petrocelis’) Mastocarpus papillatus Microcladia borealis Neorhodomela sp. Polysiphonia spp. Pyropia spp. (formerly Porphyra)

Chlorophyta Ulothrix spp. Ulva spp. Urospora spp.

Bacillariophyta Benthic diatom film Navicula spp.

Crustacea Gammarid amphipods Balanus glandula Chthamalus dalli Cirolana harfordi Pagarus sp. Idotea wosnesenskii Semibalanus cariosus

Mollusca Littorina plena Littorina scutulata Lottia digitalis Lottia paradigitalis Lottia pelta Lottia scutum Mytilus spp. Nucella caniculata Onchidella borealis Polyplacophora (chitons)

Nemertea Emplectonema gracile Polychaete

Insecta Oedoparena sp. larvae

Table F.1: Summary of species found on settlement plates, sorted by taxonomic grouping.

191