Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Abiotic and biotic processes shape species distributions and ecological communities across spatial scales Amundrud, Sarah Louise 2020

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2020_november_amundrud_sarah.pdf [ 4.23MB ]
Metadata
JSON: 24-1.0392690.json
JSON-LD: 24-1.0392690-ld.json
RDF/XML (Pretty): 24-1.0392690-rdf.xml
RDF/JSON: 24-1.0392690-rdf.json
Turtle: 24-1.0392690-turtle.txt
N-Triples: 24-1.0392690-rdf-ntriples.txt
Original Record: 24-1.0392690-source.json
Full Text
24-1.0392690-fulltext.txt
Citation
24-1.0392690.ris

Full Text

    ABIOTIC AND BIOTIC PROCESSES SHAPE SPECIES DISTRIBUTIONS AND ECOLOGICAL COMMUNITIES ACROSS SPATIAL SCALES   by   SARAH LOUISE AMUNDRUD BSc (Hons), The University of British Columbia, 2012 MSc, The University of British Columbia, 2014     A DISSERTATION 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)      AUGUST 2020  © Sarah Louise Amundrud, 2020 ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: Abiotic and biotic processes shape species distributions and ecological communities across spatial scales  submitted by Sarah Louise Amundrud  in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Zoology   Examining Committee: Diane Srivastava, Zoology Supervisor  Chris Harley, Zoology Supervisory Committee Member  Jennifer Williams, Geography University Examiner John Richardson, Forestry University Examiner   Additional Supervisory Committee Members: Amy Angert, Zoology Supervisory Committee Member Villy Christensen, Fisheries Supervisory Committee Member   iii   ABSTRACT    Species distributions and the composition of ecological communities result from the interplay of three constraints: physical barriers to dispersal, species-specific environmental requirements, and species interactions. However, how spatial scale and environmental context affect the relative importance of these constraints is still poorly understood.  I combined manipulative experiments, observational surveys along environmental gradients, and species distribution models to explore the relative importance of abiotic and biotic constraints on aquatic invertebrate communities inside bromeliads across a range of spatial scales. At the geographic scale (Central and South America), the distribution of a bromeliad-specialist damselfly was strongly limited by dispersal barriers and climate, but not by biotic interactions from other bromeliad-dwelling odonates. At the regional scale (along two elevational gradients in Costa Rica), the relative importance of abiotic and biotic constraints on species distributions and community assemblages depended on environmental context: while thermal tolerances (i.e., abiotic effects) explained invertebrate distributions under moist conditions, biotic interactions from crane fly larvae, which became predatory under drought, explained their distributions under dry conditions. At the local scale (the bromeliad), biotic interactions from crane fly larvae were disproportionally important in mediating the effects of drought and warming on community structure and ecosystem functions. This hierarchical set of studies demonstrates the scale-dependence of the interplay of abiotic and biotic processes in affecting species distributions and iv  community assemblages. While species interactions were the main drivers of community change at the local scale of the habitat, biotic effects were not important at the large geographic scale. Notably, the relative importance of abiotic and biotic processes at the landscape scale depended on environmental context, an important insight given that environmental conditions are already shifting as a result of climate change.    v   LAY SUMMARY   Species distributions and ecological communities are governed by abiotic factors like climatic constraints and physical barriers to dispersal, as well as biotic effects from other species. However, how spatial scale and environmental context affect the interplay of these constraints is still poorly understood. I combined modeling techniques with observational surveys along environmental gradients and controlled experiments to investigate the relative importance of abiotic and biotic processes on an aquatic invertebrate food web inside bromeliad plants. I found that, while biotic interactions were key drivers of community responses at small and intermediate spatial scales, they were not detectable at the large geographical scale. At the intermediate scale of the landscape, the relative importance of biotic interactions and climatic (i.e., abiotic) effects depended on environmental context, demonstrating that processes that shape ecosystems are dynamic and may shift with climate change.    vi   PREFACE   CHAPTER 2 has been previously published as:  Amundrud, S. L., Videla, M., and Srivastava, D. S. (2018). Dispersal barriers and climate  determine the geographic distribution of the helicopter damselfly Mecistogaster modesta.  Freshwater Biology 63: 214-223. DSS and MV conceived the idea. MV collected the data. I performed the analyses (except the Maxent models of potential competitors in Appendix A1, which was done by MV) and wrote the paper, with comments from DSS and MV.   CHAPTER 3 has been previously published as:  Amundrud, S. L., and Srivastava, D. S. (2020). Thermal tolerances and species  interactions determine the elevational distributions of insects. Global Ecology and  Biogeography 29: 1315-1327 I conceived the idea and designed the study with input from DSS. I and my field assistants Bret Flynn, Megan Reich, Sarina Clay-Smith, Kathleen Higgins, and Derek Wiens conducted the invertebrate survey and CTmax experiment. I analyzed the data and wrote the manuscript, with input from DSS.   vii  CHAPTER 4 has been previously published as:  Amundrud, S. L., Clay-Smith, S. A., Flynn, B. L., Higgins, K. E., Reich, M. S., Wiens,  D. R. H., and Srivastava, D. S. (2019) Drought alters the trophic role of an opportunistic  generalist in an aquatic ecosystem. Oecologia 189: 733-744 I conceived and designed Experiment 1, and SCS, BLF and MSR performed Experiment 1. I and DSS conceived and designed Experiment 2, and I, KEH and DRHW performed Experiment 2. I analyzed the data and wrote the manuscript with input from DSS. All authors provided editorial advice.  CHAPTER 5 has been previously published as:  Amundrud, S. L., and Srivastava, D. S. (2019) Disentangling how climate change can  affect an aquatic food web by combining multiple experimental approaches. Global  Change Biology 25: 3528-3538 I conceived and designed the study with input from DSS. I performed the experiment with help from Kathleen Higgins and Derek Wiens. I analysed the data and wrote the manuscript, with input from DSS.   viii   CONTENTS  Abstract ........................................................................................................................................................ iii Lay Summary ................................................................................................................................................. v Preface ......................................................................................................................................................... vi Contents ..................................................................................................................................................... viii List of Tables ................................................................................................................................................ xi List of Figures .............................................................................................................................................. xii Acknowledgements ..................................................................................................................................... xiv Dedication ................................................................................................................................................... xvi 1 Introduction .......................................................................................................................................... 1 2 Dispersal barriers and climate determine the geographic distribution of the helicopter damselfly Mecistogaster modesta ................................................................................................................................ 8 2.1 Chapter Summary ......................................................................................................................... 9 2.2 Introduction ................................................................................................................................ 10 2.3 Methods ...................................................................................................................................... 14 2.3.1 Data Collection .................................................................................................................... 14 2.3.2 Variable Selection ............................................................................................................... 17 2.3.3 Species Distribution Models ............................................................................................... 19 2.4 Results ......................................................................................................................................... 21 2.5 Discussion .................................................................................................................................... 26 3 Thermal tolerances and species interactions determine the elevational distributions of insects ..... 30 3.1 Chapter Summary ....................................................................................................................... 31 3.2 Introduction ................................................................................................................................ 32 3.3 Methods ...................................................................................................................................... 36 3.3.1 Study site and system ......................................................................................................... 36 3.3.2 Survey of bromeliad food webs .......................................................................................... 39 ix  3.3.3 Determination of insect heat tolerances ............................................................................ 40 3.3.4 Statistical Analyses .............................................................................................................. 40 3.4 Results ......................................................................................................................................... 43 3.5 Discussion .................................................................................................................................... 51 4 Drought alters the trophic role of an opportunistic generalist in an aquatic ecosystem ................... 57 4.1 Chapter Summary ....................................................................................................................... 58 4.2 Introduction ................................................................................................................................ 59 4.3 Methods ...................................................................................................................................... 62 4.3.1 Study Site and System ......................................................................................................... 62 4.3.2 Experimental manipulation of tipulid predation and water level ...................................... 63 4.3.2.1 Experiment 1: Effects of tipulids in the presence and absence of water ....................... 64 4.3.2.2 Experiment 2: Tipulid predation in high vs. low water ................................................... 66 4.3.3 Statistical analysis ............................................................................................................... 68 4.3.3.1 Experiment 1: Effects of tipulids in the presence and absence of water ....................... 68 4.3.3.2 Experiment 2: Tipulid predation in high vs. low water ................................................... 69 4.4 Results ......................................................................................................................................... 69 4.4.1 Experiment 1: Effects of tipulids in the presence and absence of water ........................... 69 4.4.2 Experiment 2: Tipulid predation in high vs. low water ....................................................... 72 4.5 Discussion .................................................................................................................................... 75 5 Disentangling how climate change can affect an aquatic food web by combining multiple experimental approaches ........................................................................................................................... 82 5.1 Chapter Summary ....................................................................................................................... 83 5.2 Introduction ................................................................................................................................ 84 5.3 Methods ...................................................................................................................................... 90 5.3.1 Study Sites and System ....................................................................................................... 90 5.3.2 Manipulative Experiment .................................................................................................... 91 5.3.3 Statistical Analyses .............................................................................................................. 94 5.4 Results ......................................................................................................................................... 95 5.5 Discussion .................................................................................................................................. 101 6 Conclusion ......................................................................................................................................... 107 Bibliography .............................................................................................................................................. 115 Appendices ................................................................................................................................................ 133 x  A Supplementary Information to Chapter 2 ........................................................................................ 134 A.1 Detailed methods and results of the Maxent models of three potential competitors of M. modesta ................................................................................................................................................ 135 A.2 References ................................................................................................................................ 138 B Supplementary Information to Chapter 3 ........................................................................................ 139 B.1  Information of the study sites on the Pacific and Atlantic slopes of Monteverde Mountain .. 140 B.2  Elevation preference and mean elevation of occurrence for bromeliad insect species for each mountainside ........................................................................................................................................ 143 B.3  Structural equation models using an obligate predator variable instead of the M. modesta variable as potential mediator to explain effects of elevation on focal species .................................. 146 C Supplementary Information to Chapter 4 ........................................................................................ 147 C.1  Analyses on the data from which all Culex spp. were removed from the total prey count ..... 148 C.2  Effects of water and cage treatment on tipulid growth in Experiment 1 ................................. 150 D Supplementary Information to Chapter 5 ........................................................................................ 151 D.1  Community composition in microcosms and natural bromeliads ............................................ 152 D.1.1 Quantification of natural bromeliad communities ........................................................... 152 D.1.2 Microcosm communities represent natural bromeliad communities .............................. 153 D.2  Evidence of supplementary prey consumption by tipulids ...................................................... 155 D.3  Treatment effects on prey emergences.................................................................................... 157 D.4  Heat tolerance (CTmax) of tipulids originating from MV and UGA ............................................. 160    xi   LIST OF TABLES  Table 2. 1: Fit of simplified “across barrier model”. ................................................................................... 21 Table 2. 2: Fit of simplified “within barrier model”. ................................................................................... 22  Table 3. 1: Abundance of five focal insect species in bromeliads as a function of elevation, mountainside, their interaction, and bromeliad capacity. ................................................................................................. 44 Table 3. 2: The abundance of each of three focal species modelled with structural equation models ..... 49  Table 4. 1: Effects of tipulids and water on mosquitoes and chironomids in Experiment 1 ...................... 70 Table 4. 2: Effects of water level and tipulids on chironomid and mosquito survival in Experiment 2 ..... 73  Table 5. 1: ANOVA results for insect survival and decomposition.............................................................. 97  Table A. 1: Environmental variables and their importance used in Maxent models of Bromeliagrion species distributions. ................................................................................................................................ 136  Table B. 1: Mountainside, elevation, geographic coordinates, and sampling dates of the nine study sites on Monteverde Mountain. ....................................................................................................................... 141  Table C. 1: Effects of tipulids and water on mosquitoes and chironomids in Experiment 1 .................... 149  Table D. 1: ANOVA results for insect emergence rates from linear models ............................................. 157      xii   LIST OF FIGURES  Figure 1. 1: Phytothelm bromeliads at Monteverde Mountain, Costa Rica ................................................. 4 Figure 1. 2: Bromeliad dwelling damselfly (Mecistogaster modesta) adult and larva ................................. 4  Figure 2. 1: Map showing location of presence and absence records of M. modesta, as well as presence records of the three potential competitor species ..................................................................................... 16 Figure 2. 2: Probability of occurrence of M. modesta for annual mean temperature (bio1) and annual precipitation (bio12).. ................................................................................................................................. 23 Figure 2. 3: The predicted realized geographic distribution of M. modesta .............................................. 24 Figure 2. 4: The predicted potential geographic distribution of M. modesta. ........................................... 25  Figure 3. 1: Diagram summarizing the study area on Monteverde mountain. .......................................... 38 Figure 3. 2: The abundances of focal species inside bromeliads surveyed along the elevation gradients on the Atlantic and Pacific slopes of Monteverde ........................................................................................... 45 Figure 3. 3: Heat tolerance (CTmax) explains elevation preference and mean elevation of occurrence (MEO) on the Atlantic slope, but not on the Pacific slope .......................................................................... 47 Figure 3. 4: Path diagram summarizing results of structural equation models for the three focal species on the Atlantic and Pacific slopes. .............................................................................................................. 50  Figure 4. 1: Effects of water and tipulids on prey community. ................................................................... 71 Figure 4. 2: Survivorship and number of surviving larvae .......................................................................... 74  Figure 5. 1: Experimental setup .................................................................................................................. 89 Figure 5. 2: Effects of drought length, Origin→Location (O→L) treatment, and tipulid presence on prey community and leaf loss ............................................................................................................................. 98 Figure 5. 3: Partial residuals showing the effect of the space-for-time treatment on prey survival.......... 99 Figure 5. 4: Partial residuals showing the effect of tipulid treatment on prey survival ........................... 100  Figure A. 1: Habitat suitability maps for potential competitor species. ................................................... 137  Figure B. 1: Sampling dates of bromeliad communities along elevational gradients............................... 142 Figure B. 2: Elevation preferences of bromeliad insects .......................................................................... 144 Figure B. 3: Mean elevation of occurrence (MEO) of each species .......................................................... 145  Figure C. 1: Effects of water and cage treatment on tipulid growth. ....................................................... 150  xiii  Figure D. 1: Natural and experimental densities of taxa in our study. ..................................................... 154 Figure D. 2: Leaf loss increases with prey mortality in communities with tipulids, but not in communities without tipulids. ........................................................................................................................................ 156 Figure D. 3:  Effects of drought length, Origin→Location (O→L) treatment, and tipulid presence on prey emergence rates ....................................................................................................................................... 158 Figure D. 4: Partial residuals showing effect of tipulid treatment on prey emergences. ......................... 159 Figure D. 5: Heat tolerance (CTmax) of tipulids. ......................................................................................... 161     xiv   ACKNOWLEDGEMENTS    This thesis was only possible because of the personal and professional support from many people. A huge thank you to my advisor, Diane Srivastava, for your insight, patience, trust in my abilities, and striking the perfect balance between letting me find my own path and providing guidance and support when needed. Thanks to my committee members Chris Harley, Amy Angert, and Villy Christensen, who have advised me on developing and thinking through my various projects.  A tremendous thank you to my research assistants: Bret Flynn, Kathleen Higgins, Derek Wiens, Megan Reich, and Sarina Clay-Smith, for your tireless efforts hiking up and down the mountain to survey bromeliads, willingness to work long days, care in tending experiments, while still finding the time to appreciate this amazing adventure and the beauty of the wilderness around us. Your help was instrumental to this work. Special thanks also to the current and past members of the Srivastava lab and the Bromeliad Working Group, especially Melissa Guzman, Natalie Westwood, Pierre Rogi, Kurt Trzcinski, Alathea Lethaw, and Andrew McDonald. I am particularly grateful to Angie Nicolas, who translated my permit applications into Spanish, and to Régis Cereghino, who was an important mentor and funder of this research.  This thesis would not have been possible without the kindness and logistical support of many people in Monteverde and elsewhere in Costa Rica. Mary and Elias Newswanger let us live and work in their own home for four months in 2015 while they “took the opportunity to travel”. Bill Haber provided advice on how to acquire the research permits, suggested study sites, and shared his love of odonates with me. Néstor Guevara and Randall Zamora Castro guided us safely through the Costa Rican wilderness, spotted the most hidden bromeliads, and shared their love and wisdom for the Costa Rican flora and fauna. Alan Pounds provided advice and shared xv  his knowledge about the natural history and the impacts of climate change on the Monteverde Cloud Forest. Special thanks also to Luis Guillermo Chaverri, who identified my mosquito larvae.   Thank you so much to the wonderful people from the Monteverde Cloud Forest Reserve and the Centro Científico Tropical (especially Yoryineth Mendez Corrales for invaluable help with logistical matters), and the Children’s Eternal Rain Forest Reserve (especially Giselle Rodriguez and Lindsay Stallcup). Thank you also to the wonderful hosts at Estación San Gerardo and Estación Pocosol for their Costa Rican hospitality and feeding us the most delicious meals. I am grateful to the land owners that allowed me to collect bromeliad invertebrates on their property: Alicia Mendes near San Luis, as well as Fred Morrison and Bill Calvert at Finca Colibri. Finally, a very big thank you to the Costa Rica Campus of the University of Georgia (especially José Montero and Fabricio Camacho), for hosting us in 2016, going through great lengths to ensure we had all the lab space and research equipment we needed, and making the second field season feel more like an extended vacation.   Last but not least, I want to thank my family. My parents Lisi and Klaus, my siblings Christoph and Laura, and my father in law Greg, thank you for supporting and believing in me. And a huge thank you to my mother in law, Linda, for the countless hours of babysitting so I could work on my PhD. Trevor, words cannot express how grateful I am for the many sacrifices you made so I could take this path, for your endless encouragement, unconditional love, for being at my side to celebrate my successes and to support me through the tough times. I truly could not have done this without you. And finally, to Cleo, for bringing a whole new kind of joy into my life, and for reminding me to once again notice the small things that no doubt sparked my interest in biology a long time ago: the snails, the worms, the dandelions, the “rollie pollies”…     xvi   DEDICATION        To Trevor and Cleo    1  CHAPTER 1   1 INTRODUCTION      2   Identifying the processes that shape species distributions and community assemblages is fundamental to our understanding of how ecosystems respond to global change. The realized niches of species are broadly defined by the interplay of three processes: dispersal limitation, habitat filtering, and species interactions (Soberón 2007). Yet, most attempts to understand species distributions and the make-up of ecological communities have focused on the effects of abiotic conditions on single species, such as physiological effects of temperature (e.g. McKechnie and Wolf 2010) and precipitation (e.g. Ernest et al. 2000; Báez et al. 2006; Ledger et al. 2008). For instance, species distribution models (SDMs) typically correlate species presence records with climatic variables, but such envelope models assume that species occur in all locations within their climatic tolerance (Kearney and Porter 2009), and thus do not take into account other limitations to species occurrence, such as altered species interactions (however, if biotic interactions systematically covary with climatic variables, climatically based models will be a good proxy for the combined effects of biotic and abiotic factors). Because of the growing amount of evidence that species interactions can be as or more important in shaping the structure of ecological communities than abiotic effects on species’ physiologies (Schmitz et al. 2003; Chase and Knight 2003; Montoya et al. 2009; Ockendon et al. 2014; Amundrud and Srivastava 2016), we need to incorporate biotic processes into models predicting the ecological effects of global change.   Abiotic and biotic processes operate at a hierarchy of spatial scales, and their relative impact on ecological communities is likely scale-dependant (Chase and Knight 2013). While biotic processes such as predation and competition often determine the composition of communities at small spatial scales (e.g., Chase and Knight 2003; Amundrud and Srivastava 2016), the relative impact of species interactions can become negligible when considering larger 3  spatial scales (Eltonian noise hypothesis: Soberón 2007). Nevertheless, even at large geographical scales, biotic interactions can be important in affecting species distributions (e.g., Heikkinen et al. 2007; Boulangeat et al. 2012; De Araújo et al. 2014). Although progress has been made in understanding the scale-dependence of the relative importance of abiotic and biotic constraints on ecological communities, this has yet to be adequately applied to understanding the effects of global change such as climate change.   For my PhD dissertation, I took advantage of the utility of bromeliad-dwelling invertebrate food webs to investigate the relative importance of abiotic and biotic processes in shaping species distributions and the make-up of communities across various spatial scales, ranging from the large continental scale to the small scale of the local habitat. By collecting rainwater and fallen detritus, tank bromeliads provide habitats for diverse communities of aquatic microorganisms and macroinvertebrates (Figure 1.1). These miniature food webs are ideal study systems to investigate how abiotic and biotic processes operate at various spatial scales, as tank bromeliads occur over landscape gradients within a large geographic range (most of the Neotropics). Bromeliad-dwelling faunal communities are relatively self-contained systems that can be easily manipulated at the local scale in order to investigate abiotic and biotic mechanisms. As these food webs have been studied extensively, we already have a detailed understanding of important processes that affect community assemblages locally. For example, I previously showed in my MSc work that a keystone predator (damselfly nymph: Mecistogaster modesta, Figure 1.2) mediates the ecological effects of drought (Amundrud and Srivastava 2016), which is expected to increase in montane areas of Costa Rica with climate change (Karmalkar et al. 2008).    4    Figure 1. 1: Phytothelm bromeliads at Monteverde Mountain, Costa Rica.         Figure 1. 2: Bromeliad dwelling damselfly (Mecistogaster modesta) adult (left) and larva (right).       5   I organized my dissertation chapters by spatial scale, going from the geographic region to the landscape to the local habitat. In Chapter 2, I employed SDMs to investigate the relative importance of physical dispersal barriers, climatic constraints, and biotic interactions in determining the distribution of the damselfly M. modesta, which spends its larval stage exclusively in bromeliads, at the geographic scale. Most SDMs assume that species occur in habitats that encompass their climatic tolerance range (Kearney and Porter 2009), and do not take into account other limitations to species occurrence, such as dispersal barriers and biotic constraints. In collaboration with Dr. Martin Videla, a former postdoc of my advisor Diane Srivastava, I incorporated dispersal constraints (the potential barriers imposed by the Andes mountain range and oceans) and biotic constraints (the presence of other bromeliad-dwelling odonate species that present potential competitors or intraguild predators) into SDMs. Model predictions suggested that at this large geographic scale, biotic factors were not important in governing the distribution of M. modesta. Instead, the distribution of this species seemed to be primarily limited by both climatic constraints (dry conditions and low temperatures) and dispersal barriers.   Next, I investigated the relative importance of abiotic and biotic processes in determining species distributions and community assemblages at the landscape scale, by surveying bromeliad food webs along two elevational gradients on the dry Pacific and wet Atlantic slopes of Monteverde Mountain, Costa Rica, as well as experimentally determining their upper thermal tolerances. These mountain slopes represent an ideal “natural experiment” to tease apart the various abiotic and biotic constraints on species regional distributions. While the ambient temperature increases towards lower elevations on both mountain slopes at roughly equal rates, the Pacific slope is significantly dryer than the Atlantic slope, allowing me to disentangle the 6  effects of temperature and precipitation on invertebrate elevational distributions. Predatory damselflies occur only at low elevations on the Atlantic slope (consistent with expectations from the SDM in Chapter 2). In Chapter 3, I investigated the relative importance of abiotic and biotic processes in driving the elevational distributions of bromeliad invertebrates. While both climate and biotic factors were important in driving invertebrate elevational distributions, their relative influences depended on environmental context. On the wet Atlantic slope, taxon-specific heat tolerances explained invertebrate elevational distributions, while biotic interactions explained their elevational distributions on the dry Pacific slope. Interestingly, it was not the damselfly that drove these biotic effects. Instead, crane fly larvae, which were previously thought to be facultative detritivores, exerted adverse effects on other invertebrates at low elevation sites on the Pacific slope (i.e., the warmest and driest sites), thus affecting their elevational distributions.   Last, I investigated the relative importance of abiotic and biotic factors in shaping bromeliad invertebrate communities at the local scale of the habitat. In Chapter 4, I investigated several potential mechanisms that may have caused crane fly larvae to exert negative effects on other invertebrates under dry environmental conditions, thus driving their elevational distributions. Through two controlled experiments, I demonstrated that drought enabled opportunistic predation by crane fly larvae via increasing predator-prey encounter rates, resulting from habitat compression due to a reduction in water volume. In Chapter 5, I combined multiple experimental approaches to test the individual and combined effects of drought, warming, and biotic interactions on community structure and ecosystem functioning. I found that adverse effects of crane flies on other invertebrates arose under drought conditions and became stronger with warming. Altered species interactions thus mediated the effects of drought and warming on the invertebrate food web at the scale of the single habitat. Indeed, the relative importance of 7  these biotic processes far outweighed any direct effects of drought and temperature on the community.  These chapters demonstrate the scale-dependence of the relative importance of abiotic and biotic processes in affecting species distributions and the make-up of ecological communities. A key insight that emerged from all chapters is that climate change will affect invertebrate communities. While indirect effects of environmental change due to altered species interactions mediated community responses at small spatial scales, such biotic effects were not detectable at large scales. Importantly, this dissertation demonstrates that the interplay between abiotic and biotic processes at intermediate spatial scales can depend on environmental context. As such, this research is shedding light on the dynamic nature of the various processes that affect ecological communities across space and time.   8  CHAPTER 2   2 DISPERSAL BARRIERS AND CLIMATE DETERMINE THE GEOGRAPHIC DISTRIBUTION OF THE HELICOPTER DAMSELFLY MECISTOGASTER MODESTA1                                                                     1 Previously published as Amundrud, S. L., Videla, M., and Srivastava, D. S. (2018). Dispersal barriers and climate determine the geographic distribution of the helicopter damselfly Mecistogaster modesta. Freshwater Biology 63: 214-223. 9  2.1 CHAPTER SUMMARY  Species’ ranges are typically constrained by the interplay of physical barriers to dispersal, environmental requirements such as suitable climatic conditions, and biotic constraints such as from predation or competition. However, teasing apart the relative importance of these constraints in determining species distributions still represents a major challenge for ecologists. The Neotropical damselfly Mecistogaster modesta (Coenagrionidae: Odonata) inhabits wet and moist forests in mainland Central America and northwestern South America. This habitat specialist spends its larval development exclusively in tank bromeliads, where it acts as a keystone predator within the aquatic food web. Although tank-forming bromeliads occur from the southern United States throughout most of South America, M. modesta is absent from the Caribbean islands and South America southeast of the Andes mountain chain. We employed species distribution models (SDMs) to explore the relative importance of physical barriers (Andes mountain range and oceanic barriers), climate (mean annual temperature and annual precipitation), and biotic interactions (competition from other bromeliad-dwelling odonates) in limiting the geographic distribution of M. modesta. We found that dispersal barriers strongly limit the geographic distribution of M. modesta. In addition, its range is restricted by low temperatures and low precipitation. Competition from other bromeliad-dwelling odonates was not important in limiting the damselfly’s range. Because of the physical barriers to dispersal, M. modesta does not occupy its full potential geographic range. Specifically, our model predicted suitable habitat on the Caribbean islands and throughout most of South America, where the species is currently absent. These findings have important conservation implications, particularly as the aridification of rainforests and subsequent localized extinctions due to climate change 10  continue. On the other hand, the species may respond to warming temperatures by tracking climate to higher elevations, with subsequent effects on naïve high-elevation bromeliad food webs. An upwards migration could also increase the probability of M. modesta overcoming the dispersal barrier presented by the Andes, enabling the damselfly to invade large areas of suitable habitat in South America.   2.2 INTRODUCTION   Damselflies and dragonflies (Odonata) are among the most ancient winged insects on Earth (~ 300 MYA: Ingley et al. 2012). As both adults and larvae typically occupy high trophic positions, these insects are often used as indicators of aquatic and terrestrial ecosystem health (Collins and McIntyre 2015). Odonates possess some of the best dispersal capabilities of all insects (Sánchez-Herrera and Ware 2012), yet these acrobatic flyers are not found everywhere. Teasing apart the relative importance of factors that determine the geographic distributions of odonates (Collins and McIntyre 2015), as well as of organisms in general (Endler 1982; De Araújo et al. 2014), still represents a major challenge for ecologists.  Major mountain ranges have led to the isolation and subsequent diversification of plants (e.g., Antonelli et al. 2009) and animals, including invertebrates (e.g., von Oheimb et al. 2013; De-Silva et al. 2016) and birds (e.g., Graham et al. 2009). Similarly, large expanses of water such as oceans and great lakes represent major dispersal barriers to most terrestrial organisms (MacDonald 2003), and can restrict the geographic ranges of even good dispersers, such as 11  winged insects (e.g., yellow-faced bumble bee: Jha 2015). Such direct physical barriers may limit the geographic distributions of odonates, and indeed, their range edges often coincide with mountain ranges and oceans (Wellenreuther et al. 2012). However, geographic barriers usually do not fully explain the geographic range limits of odonates.   The constrained geographic ranges of odonates may also be a function of their sensitivity to environmental conditions. At large spatial scales, climate, particularly temperature and precipitation, appears to ultimately determine the physiological niche of organisms (Pearson and Dawson 2003), including odonates (Collins and McIntyre 2015). Temperature determines, among other things, thermoregulation proficiency (De Marco et al. 2005), flight performance of adults (Dingemanse and Kalkman 2008; Samejima and Tsubaki 2010), development rates of larvae (Lutz 1968; Pickup and Thompson 1990; Hassall et al. 2007; Braune et al. 2008; Richter et al. 2008), time to emergence (Hassall et al. 2007; Richter et al. 2008), and voltinism (i.e., the number of generations completed within 1 year in the field, Braune et al., 2008). Precipitation, on the other hand, is strongly linked to hydroperiod in lentic waters (Collins et al. 2014) and thus determines environmental suitability for oviposition by odonates. In addition, low precipitation and associated increases in drought lengths and/or frequencies can adversely affect odonate larvae (Ball-Damerow, M’Gonigle & Resh, 2014a b; Amundrud & Srivastava, 2015, 2016).  Apart from dispersal limitation and abiotic constraints, biotic interactions such as predation and competition may also limit the distributions of odonates (Collins and McIntyre 2015). For example, the presence of predatory fish is strongly linked to the occurrence of some odonate species in ponds (McGuffin et al. 2006). In temporary waters where fish are absent, other odonate species represent strong biotic controls (e.g., Benke 1978; Johnson et al. 1985; Fincke 1994). In such systems, intraguild predation (where predators with shared prey kill and 12  eat each other: Polis, Myers & Holt, 1989), and cannibalism are the main biotic constraints (e.g., Benke 1978; Suutari et al. 2004). Biotic interactions can also limit distributions of odonates through the “ghost of competition past” (Worthen and Horacek 2015), where current geographical limits of species reflect the historical effects of competition with other organisms (Connell 1980). Although important at the local scale, biotic interactions are often considered to be negligible in affecting species distributions at large geographical scales (Eltonian noise hypothesis: Soberón and Nakamura 2009). Nevertheless, recent work suggests that biotic interactions can be important in affecting species distributions even at large spatial scales (e.g., Heikkinen et al. 2007; Boulangeat et al. 2012; De Araújo et al. 2014).  The realized geographic distributions of odonates are likely determined by the interplay of physical barriers limiting their dispersal, environmental requirements, and biotic constraints [see BAM (biotic, abiotic, and migration) diagram in Soberón, 2007]. The BAM model (Soberón 2007) posits that the range that a species fills should represent all areas within the dispersal ability of the species (M) with suitable abiotic (A) and biotic (B) conditions. Species distribution models (SDMs) can be valuable in teasing apart these factors in limiting odonate distributions (Collins and McIntyre 2015), especially if the relative contributions of the main factors conceptualized in the BAM model are well understood, or at least hypothesized (Saupe et al. 2012). However, few, if any, studies have evaluated the combined effects of climate, physical barriers, and biotic constraints on odonate ranges.  Here, we employ SDMs to explore the factors that limit the geographical distribution of the Neotropical damselfly Mecistogaster modesta (Coenagrionidae: Odonata). M. modesta oviposits its eggs exclusively in epiphytic bromeliads (Ingley et al. 2012), where the larvae feed on other insect larvae (Srivastava 2006). Adults feed on orb-weaving spiders, which they pluck 13  from webs (Hedström and Sahlén 2001; Ingley et al. 2012). Although tank-forming bromeliads occur from the southern United States (latitude ~ 30 ˚N) to almost the southern tip of South America (latitude ~ 45 ˚S; Benzing 2000), M. modesta is limited to mainland Central America and northwestern South America, inhabiting wet and moist forests from Colombia and Venezuela to Mexico-Belize (Hedström and Sahlén 2001). Throughout much of its range, M. modesta is the only odonate species associated with bromeliads, although there may be some overlap with other bromeliad-dwelling odonates in the Bromeliagrion genus towards the southern limit of its range (see Figure 2.1 and Appendix A). Within bromeliads, M. modesta nymphs have large top-down effects on the bromeliad ecosystem that eclipse the effects of any other invertebrate species. For example, M. modesta larvae have been shown to reduce abundances of other invertebrates by 25-90 % (Srivastava 2006; Hammill et al. 2015b) and decomposition by 20-50 % (Srivastava 2006; Atwood et al. 2013), increase nitrogen uptake by the bromeliad by 1200 % (Ngai and Srivastava 2006), and change CO2 flux by up to 200 % (Atwood et al. 2013; Hammill et al. 2015b). Bromeliads that contain M. modesta have fundamentally different biomass pyramids and mosquito communities than bromeliads that do not (Petermann et al. 2015; Hammill et al. 2015a). Determining which factors limit the geographic distribution of this odonate species is therefore essential to understanding spatial patterns in the structure and functioning of bromeliad food webs in Central America, and how these food webs may change under altered climate.  We used SDMs to explore the combined effects of climate, physical barriers, and biotic constraints on M. modesta. SDMs have been widely applied for freshwater species, to explore matters as diverse as assessing the impacts of climate change (e.g., benthic stream macroinvertebrates: Domisch et al. 2011), predicting suitable habitat for threatened species (e.g., 14  freshwater turtle: Costa et al. 2015), and predicting the spatial distribution of invasive species (e.g., piscivorous chub: Sato et al. 2010). A detailed review on the use of SDMs to model odonate distributions has been conducted by Collins & McIntyre (2015). We employed SDMs to reveal the relative importance of physical barriers (Andes mountain range and oceanic barriers), climate (temperature and precipitation), and biotic interactions (competition from other bromeliad-dwelling damselflies). If our model predicts no or little suitable habitat beyond the physical barriers constraining the range of M. modesta, it would indicate that climate and/or biotic constraints are the major factors limiting the damselfly’s range. In contrast, if our model predicts suitable habitat on the islands and/or southeast of the Andes, it would suggest that M. modesta is limited by dispersal barriers. As there are neither obvious physical barriers nor other bromeliad-dwelling odonates towards the northern range limits of M. modesta, we expect climatic constraints to be the limiting factors towards the north.     2.3 METHODS 2.3.1 Data Collection  We obtained occurrence records of M. modesta from several online databases, including the Global Biodiversity Information Facility (www.gbif.org), Biodiversity and Environmental Resource Data System of Belize (www.biodiversity.bz), Smithsonian National Museum of Natural History (www.mnh.si.edu), and Odonata Central (www.odonatacentral.org). We attained additional records from members of the Bromeliad Working Group 15  (http://www.zoology.ubc.ca/~srivast/bwg/researchers.html), museum curators, known experts of Odonata, and published studies on invertebrates inhabiting bromeliad phytotelmata. We also included absence records as sites where at least 20 bromeliads were inspected and no M. modesta larvae were found. When no exact coordinates were available, we estimated the geographic coordinates and the associated error according to the point-radius method (Wieczorek et al. 2004) using an online georeferencing calculator (http://manisnet.org/gc.html). We only included records with a geographical uncertainty of radius less than 5 km, resulting in a data set consisting of 101 presence and 52 absence records for M. modesta (Figure 2.1).    16    Figure 2. 1: Map showing location of presence and absence records of M. modesta, as well as presence records of the three potential competitor species. Coloured dots show occurrence records of M. modesta (green), Bromeliagrion beebeanum (blue), B. rehni (purple), and B. fernandezianum (orange). Empty dots with black crosses show M. modesta absence records.      17  2.3.2 Variable Selection  To examine the importance of dispersal in governing the distribution of M. modesta, we created a single binary “barrier” variable that indicated whether a record was collected from within vs. beyond the hypothesized dispersal barriers represented by the Pacific and Atlantic oceans and the Andes mountain range. Specifically, the layer we created defined grid cells as “beyond barriers” (value = 0) when grid cells were southeast of the summit of the Andes or more than 10 km offshore of the mainland, or “within barriers” (value = 1) if in the rest of the terrestrial study area.  To examine the importance of climate, we obtained 19 bioclimatic variables from the WorldClim database at 2.5 arc-minutes resolution (Hijmans et al. 2005), which roughly corresponds to a 5 km resolution near the tropics. To select meaningful climatic predictor variables in our model, we first reduced our data set by removing all absence records from the islands and southeast of the Andes, as we a priori hypothesized that M. modesta cannot disperse to these areas because of dispersal barriers. Including these absences could lead to erroneous predictions, as absences from habitats due to dispersal barriers, but with suitable climate, would bias a model that predicts species occurrence from climate (Peterson et al. 2011). We then estimated and ranked relative variable importance as deviance explained for the 19 climate variables by predicting M. modesta occurrence for each of the climate variables separately using logistic GLMs. Next, we examined the correlation between all possible variables to ensure that variables were not highly correlated (r > 0.8). This procedure led us to choose “annual mean temperature” (bio1) and “annual precipitation” (bio12), as these variables exhibited a high importance (they ranked 1st and 4th, respectively) and a low correlation (r = 0.50). In addition, this choice is consistent with the known physiological importance of temperature and 18  precipitation to M. modesta (Amundrud and Srivastava, unpublished data). For both climatic variables, we compared the fitted curves of two polynomial logistic models (of 1st and 2nd degree, respectively) to a smooth spline. As the simpler curves (1st degree predictor variable) represented a closer fit to the data in both cases, we used 1st degree predictor variables for bio1 and bio12 as climatic predictors of M. modesta occurrence in our final models.   Interacting species have been included as predictor variables in SDMs as either occurrence records, or as an input layer of their estimated occurrence or habitat suitability (Giannini et al. 2013; De Araújo et al. 2014). Here we implemented the latter approach to examine whether competition determines the geographic distribution of M. modesta by using the modeled distributions of three of its potential competitors: Bromeliagrion bebeanum, B. fernandezianum, and B. rehni. We focused on those species because they are present in the north of South America, so their ranges may overlap with that of M. modesta (Figure 2.1). Other bromeliad-dwelling odonate species have too few records or occur too far from the known range of M. modesta, making competition unlikely. Because there are few presence records for these Bromeliagrion species (6, 6 and 11, respectively; Figure 2.1), we performed SDMs with Maxent version 3.2.1 (Phillips et al. 2006), a widely used algorithm for presence-only data that performs better than others when the sample size is small (Hernandez et al. 2006; Wisz et al. 2008). For each competitor species, we included the habitat suitability scores obtained from the Maxent model as a continuous variable. In addition, we created variables for all possible combinations of competitor species whereby, for a given grid cell, the highest suitability value among the two (or three) competitor models was kept, resulting in seven possible competitor layers (the three competitor species individually, the three pairwise combination, and all three competitor species combined). The results were qualitatively identical for all seven competitor layers; thus, we only 19  report the results of the model in which the competition layer represents all three competitor species. This layer estimates the suitability of the habitat for at least one putative competitor of M. modesta, even if the identity of this competitor changes geographically. The detailed methods and results of the competitor models are listed in Appendix A.   2.3.3 Species Distribution Models  We employed logistic generalized linear models (GLMs) to explore the relative importance of physical dispersal barriers, climate, and competition in determining the geographic distribution of M. modesta. We chose GLMs because of their relatively high predictive power relative to other modeling approaches, as our goal was to predict M. modesta occurrence from these abiotic and biotic variables. Because it is crucial that the absence data used to train a SDM are limited to geographic areas within the dispersal ability of the species (i.e., within M in BAM model: Peterson et al. 2011), our modeling approach consisted of two steps: (1) to statistically test our expectation that the ocean and Andes represent effective dispersal barriers [and thus the absences on islands and the mainland southeast of the Andes represent records from areas outside of M (migration) in the BAM model but within A (abiotic) ∪ B (biotic)]; and, in the case that the model confirms that these absence records lie outside the area of dispersal capacity of M. modesta, (2) to predict M. modesta occurrence from climate and competition on a model trained on presence/absence records that lie within the area of dispersal capacity of the species (i.e., on the reduced data set excluding absence records on islands and south-east of the Andes, or within M in the BAM model).  20   To examine whether dispersal barriers limit the distribution of M. modesta, we employed an “across barrier” model as a logistic regression on the complete M. modesta data set (i.e., including the absence records from the islands and south-east of the Andes). As explanatory variables, we included the two climatic variables (“mean annual temperature” and “annual precipitation”), their interaction, the pooled competition variable, and the binary “barrier” variable. We used the combined competitor variable (generated from all three competitor species) as described earlier. We used stepwise regression [function “step” in R (R Core Team, 2015)] for model simplification.   Because the distribution of M. modesta was strongly limited by the oceanic and mountain barriers (see Results), we employed a second “within barrier” model that we trained on presence/absence records within the dispersal area (i.e., within M of BAM) of the damselfly. Training our model using records within M reduced the data set to 101 presences and 18 absences. As above, we used a logistic GLM to predict the probability of M. modesta occurrence from the two climatic variables, their interaction, and the combined competition variable, followed by stepwise regression.   We performed a 10-fold cross validation to check for model performance and robustness, using the R package DAAG (Maindonald and Braun 2015). To calculate threshold-dependent model performance indices, we chose the threshold to maximize Kappa using the PresenceAbsence package (Freeman and Moisen 2008) in R.   21  2.4 RESULTS   The “across barrier” analysis suggested that dispersal barriers are likely to limit the geographic distribution of M. modesta, as the barrier variable, as well as the two climatic variables (bio1 and bio12), were significant predictors of the damselfly’s occurrence (Table 2.1a). The competition variable and the interaction term of bio1 x bio12 were removed from the model by stepwise regression. The total variance explained by the model was 71.6%, and 10-fold cross validation established model robustness (Table 2.1b).    Table 2. 1: Fit of simplified “across barrier model” to the full dataset of presence-absence records. Data to train the model included absences outside the hypothesized area of dispersal. The initial model included dispersal barriers, climatic constraints [annual mean temperature (bio1) and annual precipitation (bio12), as well as their interaction] and biological constraints (competitors). Competition and the bio1 x bio12 interaction terms were removed from the model by stepwise regression. (a) ANOVA table after stepwise regression. (b) Model performance indices and 10-fold cross validation (CV). (a) DF Deviance Residual DF Residual Deviance p-value NULL   152 196.13  bio1 1 13.57 151 182.56 < 0.001 bio12 1 12.89 150 169.67 < 0.001 barrier 1 114.05 149 55.62 < 0.001    (b) Threshold PCC Sensitivity Specificity Kappa AUC TSS Model 0.45 0.95 0.99 0.87 0.88 0.97 0.86 10-fold CV 0.45 0.93 0.97 0.85 0.84 0.96 0.82   22   The “within barrier” analysis suggested that climatic constraints are likely important limiting factors of M. modesta occurrence within the potential dispersive area of the species (Table 2.2a). As in the “across barrier” model, the interaction of bio1 x bio12 and the competition variable were removed from the model by stepwise regression. The total variance explained by the final model was 45%, and a 10-fold cross validation showed adequate model robustness (Table 2.2b). A closer examination of the climatic constraints on M. modesta occurrence revealed that the species is limited by low temperatures and dry conditions, as rises in annual mean temperature (bio1) and annual precipitation (bio12) significantly increase the predicted probability of M. modesta occurrence (Figure 2.2). Specifically, the mean annual temperature and precipitation for which our model predicts a 50% probability of M. modesta occurrence is 16.22˚C (± 0.95 ˚C SE), and 1091 mm (± 217 mm SE), respectively (Figure 2.2).  Table 2. 2: Fit of simplified “within barrier model” to the reduced dataset of presence-absence records. Data to train the model did not include absences outside the hypothesized area of dispersal. The initial model included climatic constraints [annual mean temperature (bio1) and annual precipitation (bio12), as well as their interaction] and biological constraints (competitors). Competition and the bio1 x bio12 interaction terms were removed from the model by stepwise regression. (a) ANOVA table after stepwise regression. (b) Model performance indices and 10-fold cross validation (CV). (a) DF Deviance Residual DF Residual Deviance p-value NULL   118 101.12  bio1 1 36.44 117 64.69 < 0.001 bio12 1 9.07 116 55.62 0.002   (b) Threshold PCC Sensitivity Specificity Kappa AUC TSS Model 0.57 0.92 0.97 0.67 0.68 0.93 0.64 10-fold CV 0.57 0.91 0.96 0.61 0.61 0.85 0.57  23   Figure 2. 2: Probability of occurrence of M. modesta for (a) annual mean temperature (bio1) and (b) annual precipitation (bio12). Empty circles represent presence (at 1) and absence (at 0) records of M. modesta. Solid lines represent the predicted probabilities of M. modesta occurrence from logistic regression models, and dashed lines represent their 95 % confidence intervals.     The predicted realized geographic distribution of M. modesta encompasses most of Central America (except at high elevations) and the northern parts of South America, spanning from southern Mexico below ~ 20˚N (and further north at both of Mexico’s coasts) to the northeastern side of the Andes mountain range in South America (from east Venezuela to Ecuador and possibly into Peru; Figure 2.3). However, the potential geographic distribution predicted by our model spans far into South America to ~ 30˚S and includes the Caribbean islands, suggesting that suitable habitat for M. modesta exists beyond the physical barriers represented by the ocean and Andes mountain range (Figure 2.4).  24    Figure 2. 3: The predicted realized geographic distribution of M. modesta (white to green colour scale indicates probability of occurrence of M. modesta). This species distribution model is restricted to areas that are within the dispersal ability of the damselfly.      25     Figure 2. 4: The predicted potential geographic distribution of M. modesta (white to green colour scale indicates probability of occurrence of M. modesta). Although the damselfly is not currently present on islands and east of the Andes, our model predicts that the climate there is suitable for the damselfly.     26  2.5 DISCUSSION   Our analysis suggests that physical barriers to dispersal and climatic constraints likely determined the distribution of M. modesta in our study, but competition from other bromeliad-dwelling odonates was not important. In particular, we identified two dispersal barriers: oceanic distances in excess of 10 km and the Andes mountain chain, which likely exclude M. modesta from the Caribbean islands and from much of South America. In addition, annual mean temperature and annual precipitation were likely also key factors limiting the geographic distribution of M. modesta. These climatic constraints excluded M. modesta from high elevations (too cold), as well as from latitudes above its northern range edge in Mexico (too dry) and from west of the Andes (too dry and/or too cold; Figure 2.3). As dispersal barriers exclude M. modesta from suitable habitat on the Caribbean islands and southeast of the Andes (Figure 2.3), our study indicates that this damselfly does not occupy its full potential geographic distribution (Figure 2.4).  M. modesta is unlikely to occupy its full potential geographic niche, as our “across-barrier” analysis predicted suitable habitat on the Caribbean islands and southeast of the Andes. We conclude that M. modesta, like other Neotropical animals (Emmons and Feer 1997), has not been able to cross the ocean expanse between the Central American mainland and the Caribbean islands to fill its potential range, consistent with marine barriers limiting animal distributions (MacDonald 2003; Smith 2009), even for taxa with good dispersal capabilities such as winged insects (Jha 2015). In addition, the geographic distribution of M. modesta is constrained by the Andes mountain range, which excludes it from suitable habitat in much of South America (Figure 2.4). This is not surprising, as the Andes represent a significant barrier for many 27  organisms including plants (e.g., Antonelli et al. 2009; Givnish et al. 2016), insects (e.g., De-Silva et al. 2016), and birds (e.g., Graham et al. 2009). It is likely that this mountain range represents a barrier to M. modesta because of unfavourable environmental conditions at high elevations, and indeed, Odonata richness often decreases with increasing elevations (e.g., Ecuadorian Andes: Jacobsen, 2003; Amundrud and Srivastava, pers. obs.). However, our analysis does suggest that potentially suitable habitat corridors exist through the Andes, notably in east-central Columbia and northwest Venezuela (Figure 2.4). It is possible that M. modesta has not successfully migrated through these low elevation corridors simply because the habitat conditions there are not suitable despite a suitable climate (e.g., bromeliads or primary forest may be absent). Behavioural tendencies of M. modesta to avoid open areas may be more important than its ability to disperse, consistent with the discovery that dispersal behaviour is more important in predicting odonate range size than is dispersal capacity (McCauley et al. 2014). Indeed, biological corridors are also inefficient for the giant tree-hole breeding damselfly Megaloprepus caerulatus, a species related to Mecistogaster modesta, which is sensitive to habitat fragmentation and does not travel through open areas (Khazan 2014). However, it is important to note that odonates in the Bromeliagrion genus that do fulfill the bromeliad niche east of the Andes may occur across these physical barriers (Appendix A), although too few records exist to make accurate predictions of their exact ranges. Any behavioural constraints of M. modesta that impede it from passing though such corridors may thus not apply to bromeliad-dwelling odonates in general.   In addition to physical barriers, climatic constraints were important factors limiting the geographic range of M. modesta, excluding the species from cold and dry areas. Although temperature and precipitation together affect water levels in bromeliads (i.e., high temperatures 28  and low precipitation lead to the driest conditions in bromeliads), we did not find a significant interaction of precipitation and temperature (the interaction term was removed from the models by stepwise regression). Furthermore, damselflies were restricted by low temperatures rather than the high temperatures that would be associated with faster evaporation rates. Instead, both annual mean temperature and annual precipitation had strong individual positive effects on the probability of M. modesta occurrence (Figure 2.2). Direct positive physiological effects of temperature on M. modesta thus appear to be more important than indirect negative effects of temperature by affecting the hydroperiod of bromeliads, consistent with studies showing that odonate occurrence (Ball-Damerow et al. 2014a) and richness (Jacobsen 2003) increases with higher temperatures. Our finding that low precipitation limited M. modesta occurrence is also not surprising, as experimental evidence suggests that M. modesta larvae are disproportionately affected by drought (Amundrud and Srivastava 2015, 2016), which results from low precipitation in bromeliads (Zotz and Thomas 1999). Other studies have also found negative effects of low precipitation and drought on odonates (e.g., Ball-Damerow et al. 2014a, b), particularly for those species that are habitat specialists (Ball-Damerow et al. 2014a).   Our analysis suggests that biotic factors from potential competitors were likely not important in limiting the distribution of M. modesta. However, the competition layer used in our models correlated with the layer representing dispersal barriers, and such correlations can result in biotic interactions being masked (Godsoe et al. 2017a). Still, closely related species of giant damselflies that breed in water-filled tree holes do coexist despite strong intraguild predation, because large larvae kill smaller ones independently of species identity, ensuring that competitive dominance is more a function of larval size than species identity (Fincke 1994). Similarly, our study shows that in South America, the ranges of the three bromeliad-dwelling 29  damselflies (Bromeliagrion bebeanum, B. fernandezianum, and B. rehni) do overlap (see Appendix A), suggesting that competition does not constrain the distributions of those species at a large spatial scale.   To summarize, both physical barriers to dispersal and climatic constraints were important limiting factors of the geographic distribution of M. modesta, while competition from other bromeliad-dwelling odonates was not. Our finding that the distribution of this damselfly is strongly linked to climate has important conservation implications, particularly as aridification of rainforests due to climate change continues (Brodie et al. 2012). Conversely, this species may benefit from rising temperatures by tracking climate to higher elevations. The introduction of this novel predator to high elevation systems would have strong effects on naïve high-elevation bromeliad food webs. An upwards migration by M. modesta could also increase the probability of this species overcoming the dispersal barrier presented by the Andes, enabling it to invade large areas of suitable habitat in South America which are currently beyond reach. Such a range expansion from climate change has happened, for example, for mountain pine beetles in British Columbia, Canada (Carroll et al. 2003). Future research should consider how simultaneous changes in precipitation and temperature in the Neotropics can alter the constraints on keystone species such as this damselfly.    30  CHAPTER 3   3 THERMAL TOLERANCES AND SPECIES INTERACTIONS DETERMINE THE ELEVATIONAL DISTRIBUTIONS OF INSECTS2                                                               2 Previously published as Amundrud, S. L., and Srivastava, D. S. (2020) Thermal tolerances and species interactions determine the elevational distributions of insects. Global Ecology and Biogeography 29: 1315-1327 31  3.1 CHAPTER SUMMARY   While physiological limits to thermal extremes are often thought to determine the abundance and geographic distribution of species, more recent evidence suggests that species interactions may be equally important. Moreover, the relative importance of these constraints may shift with changing abiotic conditions, such as climate change. Here, we explore the relative importance of physiological tolerances to heat and species interactions in determining the distribution of insects along two elevational gradients. The gradients contrast in precipitation but not temperature, allowing us to separate these two climatic factors. We estimated the elevation preferences of the aquatic larvae of five bromeliad-dwelling insect taxa by surveying 170 bromeliads along the moist Atlantic and the dry Pacific slopes of Monteverde, Costa Rica, and experimentally determined their critical thermal maxima (CTmax). We determined if taxon-specific heat tolerances predict their elevation preferences, using Deming regressions, and tested if potential predators mediated elevation effects on insect distributions, using structural equation models. On the moist Atlantic slope, heat tolerances of insects explained their elevational distributions: taxa with high heat tolerances preferred low elevations where conditions are warmest, while taxa with low heat tolerances preferred high elevations where it is coldest. By contrast, on the drier Pacific slope, the elevational abundance pattern of many insects reflected negative interactions from crane fly larvae. These larvae are known to become predatory under drought conditions and were disproportionally abundant at low elevations on the Pacific slope. In summary, we show that under drought, indirect effects mediated by species interactions can override any direct physiological effects of environmental conditions on insect distributions. The relative importance of limits to physiological tolerance and species interactions thus depends on 32  environmental context, an important insight given that environmental conditions are expected to shift with climate change.   3.2 INTRODUCTION   In a rapidly changing world, understanding how species distributions respond to changes in abiotic and biotic conditions is becoming ever more important. The abiotic niches of species can be broadly defined by their tolerance to climatic conditions such as temperature and precipitation. Species physiological limits to thermal extremes have often been considered to have major controls on their abundances and geographic distributions (e.g., Gaston and Chown 1999; Bozinovic et al. 2011; Kellermann et al. 2012). More recent evidence, however, suggests that species interactions may be at least as important in determining species distributions as physiological constraints of organisms (e.g., Ockendon et al. 2014; Godsoe et al. 2017b). Yet, the interplay of abiotic and biotic factors in shaping species distributions and abundances, and thus spatial variation in the composition of ecological communities, is still poorly understood.   Temperature is one of the most important abiotic constraints on the distribution and abundances of living organisms (Sunday et al. 2011, 2012), resulting from its profound effects on biological functions at all levels of biological organization (Hochachka and Somero 2002). As predicted by the thermal adaptation hypothesis (Sunday et al. 2012), thermal tolerance limits of ectotherms often correlate with species distributions (Addo-Bediako et al. 2000; Kellermann et 33  al. 2012; Vorhees et al. 2012). Much of the evidence supporting the thermal adaptation hypothesis comes from studies testing thermal tolerance limits in ectotherms along latitudinal (e.g., Addo-Bediako et al. 2000; Hoffmann et al. 2002) and elevational (e.g., Gaston and Chown 1999; Oyen et al. 2016) gradients. Tolerance to extreme heat events is often considered to be the most important predictor of species distributions (Clusella-Trullas et al. 2011; Kellermann et al. 2012; Vorhees et al. 2012; Overgaard et al. 2014; Kaspari et al. 2015), and studies often use measures of “heat tolerance”, which identify the upper bounds of species’ thermal tolerances, as an ecologically relevant method to link the thermal limits of species to their distributions (e.g., Cooper et al. 2008; Jørgensen et al. 2019). However thermal tolerance cannot always adequately explain species distributions (e.g., Nowrouzi et al. 2018).   Species interactions, by affecting local abundance, may also constrain the geographic distributions of species (Ockendon et al. 2014), but there are mixed views on the importance of this constraint (Nicholson 1958; Davis et al. 1998; Silvertown 2004). Species can be influenced by interactions from other species from the same, higher, or lower trophic levels, and interactions can range from inhibitory (Lawton and Hassell 1981) to beneficial (Bruno et al. 2003). Importantly, the strength and overall importance of species interactions can vary with changing abiotic conditions (Chase and Knight 2003; Hein et al. 2013; Blois et al. 2013; Ockendon et al. 2014; Amundrud and Srivastava 2016; Godsoe et al. 2017b; Amundrud and Srivastava 2019; Amundrud et al. 2019). For example, the magnitude of top-down effects exerted by consumers is often temperature dependent (Marino et al. 2018), as temperature affects metabolic rates (Vucic-Pestic et al. 2011) and thus alters consumption rates in aquatic (e.g., damselflies: Thompson 1978; fish: Meerhoff et al. 2007) and terrestrial (e.g., spiders: Barton and Schmitz 2009; beetles: Vucic-Pestic et al. 2011) systems. Similarly, changes in hydrological regimes in aquatic systems, 34  which depend on rainfall and evaporation rates (Schindler 2001), can impose biomechanical or metabolic constraints on organisms (Woodward et al. 2016) and thus alter species interactions (Chase and Knight 2003; Amundrud and Srivastava 2016, 2019; Amundrud et al. 2019). Due to the mediating role of climate in species interactions, the relative importance of physiological tolerance limits and species interactions in controlling species distributions and abundances likely depends on environmental context.  Surveys of communities along environmental gradients have been critical in developing our mechanistic understanding of the importance of biotic and abiotic regulators of species distributions. Altitudinal gradients in particular have been useful in improving our knowledge of the drivers of species distributions, as considerable climatic variation exists over short distances, diminishing any effects of dispersal limitation that can confound latitudinal gradients (Qian and Ricklefs 2007). As such, any turnover in community composition along altitudinal gradients is likely driven by changing abiotic conditions and biotic factors (Carpenter 2005). Yet, it has been challenging to disentangle which climatic factors are driving distributional limits of species, as multiple climatic factors (such as temperature, precipitation, solar radiation, etc.) change in concert.   Here, we present the results of an observational study exploring the relative importance of heat tolerance and species interactions in determining the elevational distributions of aquatic insects inhabiting phytothelm bromeliads. We contrast two elevation gradients, which span the Pacific and Atlantic slopes of the mountain of Monteverde, Costa Rica, providing us with an ideal opportunity to tease apart the effects of temperature from the effects of precipitation on the communities: while temperature increases at equal rates towards lower elevations on both the Pacific and Atlantic slopes, the Atlantic slope is relatively wet, while the Pacific slope is 35  relatively dry, especially towards lower elevations (Figure 3.1). Bromeliads and their associated macroinvertebrate communities occur naturally along these gradients, and bromeliad food webs are known to be strongly governed by species interactions. These include positive interactions from detritivorous crane fly (Trentepohlia spp.) larvae on other insects through detrital processing (Starzomski et al. 2010), as well as negative top-down effects from predatory damselfly (Mecistogaster modesta) nymphs on the communities (Amundrud and Srivastava 2016). Importantly, crane fly larvae have recently been shown to become predatory under drought conditions (Amundrud and Srivastava 2019; Amundrud et al. 2019), while the persistence of damselfly nymphs has been shown to be limited by drought (Amundrud and Srivastava 2016).  A central prediction of the thermal adaptation hypotheses is that tropical insect distributions are tightly linked to their physiological tolerance of thermal extremes (Deutsch et al. 2008; but see Vasseur et al. 2014). We predicted that if thermal tolerance limits are the main drivers of the elevational distributions of insects, then taxa with low tolerances to heat will be distributed towards higher elevations where conditions are coldest, while taxa with high tolerances to heat will be more resistant to warmer conditions and thus can thrive at lower elevations on both mountain slopes (Prediction 1). However, if species interactions are the primary driver of insect elevational distributions, we predicted that top-down effects from damselflies and/or crane flies will explain elevational distributions better than heat tolerance (Prediction 2). Finally, if the relative importance of species interactions depends on environmental context (i.e., drought), the relative influence of species interactions and temperature in driving insect distributions will depend on mountainside (Prediction 3). Such context-dependence could occur if negative interactions from damselflies shape insect 36  distributions on the Atlantic slope, but not on the Pacific slope, which may be too dry for damselflies to persist (Prediction 3.1). Alternatively, negative interactions from crane flies could shape the elevational distributions of other taxa on the Pacific slope, where drought enables crane flies to become predatory, but not on the Atlantic slope which does not experience drought conditions (Prediction 3.2).    3.3 METHODS 3.3.1 Study site and system  The study area extended from the high elevation cloud forest of Monteverde, northwestern Costa Rica (10°18'N 84°46'W, highest elevation ~ 1600 m), and down the Atlantic (10°21'N 84°40'W, lowest elevation ~ 800 m) and Pacific (10°17'N 84°48'W, lowest elevation ~ 1100 m) slopes of the Tilaran mountain range (Figure 3.1). All study sites were in the Monteverde Cloud Forest Reserve and the Children’s Eternal Rain Forest Reserve, except for the lowest elevation site on the Pacific slope, which was on private land. Although we attempted an equal sampling effort at both mountainsides, this was not possible, as bromeliads were rare on the dry Pacific slope and much of the forest has been converted to farm land. Elevational gradients in the tropics exhibit particularly high spatial turnover in temperature (Dillon et al. 2006; García-Robledo et al. 2016). While the mean annual temperature increases at roughly equal rates towards lower elevations on both mountain sides, elevations below ~ 600 m on the Pacific slope are experiencing significantly greater temperature variation (defined as Bioclim 37  1960-1990 estimated temperature annual range) than the same elevations on the Atlantic slope (Baudier et al. 2018). However, over the elevational extent of our study (> 800m), both mountain slopes experienced similar elevation-related temperature changes: mean and maximum daily temperature (recorded by temperature loggers from Sep. 9-30, 2016; see legend of Figure 3.1 for more information) increased by ~ 6°C and ~ 7.5°C, respectively, per 1000 m drop in elevation on both mountain slopes (Figure 3.1a). The two slopes do however differ in humidity, with low elevations on the Pacific slope being significantly drier than low elevations on the Atlantic slope (Figure 3.1b). Bromeliads occur naturally along these elevational gradients and exhibit variation in thermal and hydrological regimes according to local environmental conditions. Tank bromeliads at Monteverde [Werauhia gladioliflora (H. Wendl.) Antoine, Werauhia sanguinolenta Cogn. and Marchal, and Guzmania scherzeriana Mez] contain detrital based food webs consisting of macroinvertebrates and microorganisms. The dominant macroinvertebrates inside bromeliads are the larvae of dipterans (filter-feeding Culicidae: mainly Culex spp., Wyeomyia spp., and Anopheles spp.; detritivorous Chironomidae: especially Polypedilum spp. and Orthocladiinae; and omnivorous Tipulidae: undescribed Trentepohlia spp.), as well as detritivorous coleopteran Scirtidae (Scirtes spp.). At elevations below ~ 1300 m on the Atlantic slope, predatory damselfly larvae Mecistogaster modesta Selys. (Coenagrionidae: Odonata) occur at low abundances, and predatory larvae and adults of a diving beetle (Dytiscidae: Copelatus spp., Coleoptera) are common at the lowest elevation site (Figure 3.1).    38  Figure 3. 1: Diagram summarizing the study area on Monteverde mountain. The main diagram depicts the elevations and sampling effort (i.e., number of bromeliad sampled: n) of the nine study sites, as well as the species composition at each site. Line drawings represent insect species (i.e., the larvae of the mosquito Culex erethyzonfer, the chironomid Polypedilum spp., the marsh beetle Scirtid spp., the crane fly Trentepohlia spp., the damselfly M. modesta, and the diving beetle Copelatus spp.). Filled circles represent the relative abundances* of the species at each study site. Colour bars along the mountain represent temperature (Atlantic slope = Pacific slope) and humidity (Atlantic slope ≠ Pacific slope) change along elevation. Subset figures on top right show (a) the average daily mean (marks and solid/dashed lines), maximum and minimum (dotted lines) temperature†, and (b) the mean daily humidity (±95% C.I.) obtained from weather stations (Feb. 2 – May 15, 2013) on the Atlantic (green circles and solid line) and Pacific (purple crosses and dashed line) slopes. The lines in (b) were fit with logistic regression models and are for visual purposes only (i.e., these model were not used in further analyses).   *For each species, its average abundance was calculated from sites where at least one individual of that species was present (e.g., for M. modesta, only the four lowest elevation sites at the Atlantic slope were included in the calculation, as this species was absent from all other sites). 39  † We placed iButtonTM data loggers (DS1921G‐F5# Thermochron, 4K) inside and outside (tied on and shaded by the plant) bromeliads, which recorded the temperature hourly from Sep. 9-30, 2016, the timeframe for which we had overlapping temperature data for all loggers for all sites. Habitat temperature did not depend on mountainside or whether or not temperature loggers were placed inside or outside bromeliads.    3.3.2 Survey of bromeliad food webs  We surveyed bromeliad food webs at nine sites along the two elevation gradients (Figure 3.1) from late September to early December in 2015 and 2016 (see Appendix B.1 for names, elevations, exact coordinates and sampling dates of sites). Because sampling time can affect abundance patterns of insects due to seasonal effects, we ensured that sampling was completed in the shortest possible timeframe, and performed non-systematically with respect to mountainside and elevation (see Figure B.1 in Appendix B.1). We quantified the aquatic macroinvertebrate communities inside bromeliads with a water-holding capacity of up to 1000 ml (n = 14 – 25 depending on site: Figure 3.1) by rinsing the entire contents of bromeliads, searching bromeliad water and detritus thoroughly for insects and other invertebrates, and recording the identities (finest taxonomic level or morphospecies) and abundances of all organisms  > 1 mm in length. We measured the water-holding capacity (in mL) of each bromeliad by adding a known quantity of water to the empty bromeliads until overflowing, and then subtracting the overflow from the volume added. We also quantified the detritus inside bromeliads as the dry weight of leaf fragments retained by a sieve with a mesh size of 850 µm. We transported live insect larvae collected in 2015 back to the laboratory in Santa Elena, where they were maintained at constant room temperature of ~ 20 °C until they were used in further experiments. 40  3.3.3 Determination of insect heat tolerances  We experimentally determined the heat tolerances of five common macroinvertebrate taxa [n = 10 – 12 individuals per taxon per site of the larvae of the mosquito Culex erethyzonfer (filter-feeder), the chironomid Polypedilum spp. (collector), the marsh beetle Scirtid spp. (scraper), the crane fly Trentepohlia spp. (shredder), and the damselfly M. modesta (predator)]. Larvae were collected in 2015 at two sites on the Atlantic slope (at 1379 m and 1221 m elevation) and two sites on the Pacific slope (at 1517 m and 1082 m elevation; Figure 3.1). We determined heat tolerances as the critical thermal maxima (CTmax) of larvae that have been acclimatized at the laboratory for at least 4 days (Brattstrom 1968; Duarte et al. 2012) by subjecting individuals to gradually increasing temperatures (0.3 °C min-1) in an electric water bath. Such dynamic assays of temperature ramping have been found to be suitable in characterizing heat tolerance in insects (Jørgensen et al. 2019). We estimated heat tolerance as the temperature at which the insect first failed to respond to a tactile stimulus (e.g., Buchanan et al. 1988; Kivivuori and Lahdes 1996; Cooper et al. 2008; Jørgensen et al. 2019).  3.3.4 Statistical Analyses  We performed all analyses in R (version 3.6.2: R Core Team 2019). From our total data set of 170 bromeliads, we removed outliers, defined as bromeliads with disproportionately high macroinvertebrate abundances totalling > 400 individuals (n = 4), resulting in a final data set consisting of 166 bromeliads (see Figure 3.1 for sample size at each study site). We explored the potential effects of elevation, mountainside, or their interaction on each of bromeliad size and detrital content, using linear models. To quantify the effect of elevation, mountainside, and their 41  interaction on abundances of focal taxa, we employed generalized linear models (GLMs) with a Poisson error distribution. In these insect abundance models, we included bromeliad capacity as covariate to control for any effects of bromeliad size that could confound effects of elevation and/or mountainside. We examined model suitability by checking for overdispersion in Poisson models, as well as by using quantile–quantile, residual, and outlier plots. We employed the glm.poisson.disp function (R package dispmod: Scrucca 2018) to correct for overdispersion in Poisson models.  We estimated the elevation preference (EP) of each taxon by extracting the slope from the elevation term of the Poisson model. A positive slope thus indicates a preference for high elevations, while a negative slope indicates a preference for low elevations. When the interaction of mountainside with elevation was not significant (Scirtid spp. and Polypedilum spp.), we extracted the slope of a simplified model that did not include mountainside. When mountainside interacted with elevation (C. erethyzonfer and Trentepohlia spp.), we fit separate models for the Pacific and Atlantic mountainside, and extracted the slope from each model.   As an alternative measure of insect differences in sensitivity to elevation, we also calculated the mean elevation of occurrence of each taxon for each mountainside as the average elevation of occurrence over all individuals within each taxon. We stress that, while this metric allows us to compare elevation midpoints of various taxa within this survey data, any comparisons to other data sets (or indeed between the two mountainsides) are not appropriate, as this metric is strongly influenced by sampling extent and unequal sampling effort across elevation. 42   To test Prediction 1 (thermal tolerance limits are the main drivers of elevational insect distributions) and Prediction 3 (the relative importance of thermal tolerance and species interactions depends on environmental context), we investigated the relative importance of heat tolerance on the EP of bromeliad insects for each side of the mountain by employing a Deming regression in the deming R package (Therneau 2018). Deming regression is suitable for regressions between variables both measured with error, and both heat tolerance and EP were estimates with standard errors. To test the robustness of results based on EP, we also performed a weighted regression [weights = 1/(CTmax^2)] in which we used the mean elevation of occurrence instead of EP as response.   To test Prediction 2 (species interactions are the primary driver of insect elevational distributions) and Prediction 3 (the relative importance of thermal tolerance and species interactions depends on environmental context), we employed structural equation models (SEMs), using the lavaan package in R (Rosseel 2012), to explore how elevational distributions of insects were related to direct effects of elevation versus indirect effects mediated by the potential predators M. modesta (Prediction 3.1) and Trentepohlia spp. (Prediction 3.2). Although we do not measure species interactions per se (this would require measurements of per capita interaction strength), we predicted that predation would result in a negative correlation of the relative abundances of potential predators with potential prey species, as both M. modesta and Trentepohlia spp. have been shown in previous experiments to exert negative per capita interactions on other insects (e.g., Amundrud and Srivastava 2016; Amundrud et al. 2019). However, we do acknowledge that the relative abundance of a co-occurring species could also reflect shared (unmeasured) habitat characteristics (Araújo and Luoto 2007). To meet the assumption of equal variances in SEMs, we recoded the terms in our model by dividing elevation 43  by 150 and multiplying the relative abundances of our focal taxa by integers ranging from 5 (for Scirtid spp. and Polypedilum spp., the most common taxa) to 100 (for M. modesta, the rarest taxon) so that variances were within one order of magnitude of one another (Rosseel 2012).    3.4 RESULTS  Bromeliads tended to become smaller towards higher elevations (F1,162 = 29.9, p < 0.001), but bromeliad size did not depend on mountainside (F1,162 = 1.77, p = 0.18) or the interaction of elevation with mountainside (F3,162 = 0.88, p = 0.35). Detrital content inside bromeliads did not depend on elevation (F1,160 = 1.04, p = 0.31), mountainside (F1,160 = 3.58, p =0.06), or their interaction (F1,160 = 0.41, p = 0.52).  Bromeliad-dwelling insect taxa exhibited distinct elevation preferences (Figure 3.2 and Appendix B.2). The larvae of mosquitoes (C. erethyzonfer) and marsh beetles (Scirtid spp.) both increased in abundance towards higher elevations, although for mosquitoes this preference was only apparent on the Atlantic slope (Table 3.1, Figure 3.2a,c). By contrast, the predatory damselfly nymph (M. modesta) and the facultative predatory crane fly larva (Trentepohlia spp.) preferred lower elevations; specifically, damselfly abundance increased towards low elevations on the Atlantic slope (this species was absent from the Pacific slope and from elevations > ~1250 m on the Atlantic slope), whereas crane fly abundance was highest at the low elevation site on the Pacific slope (Table 3.1, Figure 3.2d,e). The chironomid larva (Polypedilum spp.) was the only taxa whose abundance was not affected by elevation. However, this insect was twice as 44  common on the Atlantic slope as on the Pacific slope (Table 3.1, Figure 3.2b). Bromeliad capacity only affected the abundances of two taxa, marsh beetles and chironomids, which both increased in abundance with bromeliad size (Table 3.1). The mean elevation of occurrences of bromeliad insects mirrored their elevation preferences (Appendix B.2).     Table 3. 1: Abundance of five focal insect species (“taxon”) in bromeliads as a function of elevation (continuous), mountainside (categorical: Atlantic vs. Pacific), their interaction, and bromeliad capacity (continuous). For the damselfly M. modesta, mountainside is not included, as this species was only found on the Atlantic slope. For each focal taxon, all explanatory variables were fit in a single Poisson generalized linear model, with terms assessed with an ANOVA as indicated by Χ2 and p-values.  elevation mountainside elevation:side bromeliad capacity Taxon Χ2 p Χ2 p Χ2 p Χ2 p C. erethyzonfer 13.07 <0.001 1.27 0.259 4.34 0.037 0.21 0.647 M. modesta 6.13 0.013 - - - - 2.91 0.088 Polypedilum spp. 0.05 0.832 17.43 < 0.001 2.69 0.101 7.24 0.007 Scirtid spp. 14.51 <0.001 2.07 0.150 0.21 0.646 31.04 <0.001 Trentepohlia spp. 7.90 0.005 7.37 0.007 13.18 <0.001 0.56 0.454     45  Figure 3. 2: The abundances (abd.) of focal species inside bromeliads surveyed along the elevation gradients on the Atlantic (green circles, solid lines) and Pacific (purple crosses, dashed lines) slopes of Monteverde. Focal species are the larvae of (a) the mosquito Culex erethyzonfer, (b) the chironomid Polypedilum spp., (c) the marsh beetle Scirtid spp.,(d) the crane fly Trentepohlia spp., and (e) the damselfly M. modesta. Lines represent best fit from Poisson regressions models. When mountainside interacted with elevation (C. erethyzonfer and Trentepohlia spp.) or when only mountainside was statistically significant (Polypedilum spp.), separate lines are shown for the Pacific and Atlantic slopes (purple, dashed or green, solid); otherwise, the regression line from the model without mountainside is shown (Scirtid spp.: black, longdashed). M. modesta abundance was only modeled on the Atlantic slope, as it was absent from the Pacific slope.   46  Bromeliad insects also exhibited distinct heat tolerances. Mosquitoes had the lowest tolerance to heat (CTmax = 38.5°C  ± 0.3 (SE), n = 41), followed by chironomids (CTmax = 40.5°C  ± 0.3 (SE), n = 42). The heat tolerances of marsh beetles (CTmax = 41.9°C  ± 0.2 (SE), n = 43) and crane flies (CTmax = 42.2°C  ± 0.3 (SE), n = 40) were similar, and damselflies exhibited the highest heat tolerance of all bromeliad insects in our study (CTmax = 44.9°C  ± 0.3 (SE), n = 12). Heat tolerances did not differ between mountainsides, so we only report the combined (i.e., average) CTmax for each taxon. Taxon-specific heat tolerances were related to elevation preferences and mean elevation of occurrences of bromeliad insects on the Atlantic, but not the Pacific slope of Monteverde (Figure 3.3). On the Atlantic slope, insect taxa that exhibited high heat tolerances tended to disproportionally occur at low elevations - where it is warm - while taxa with low heat tolerances tended to disproportionately occur at high elevations - where it is cold (Deming regression: p = 0.002, Figure 3.3a; weighted regression: F1,3 = 10.5, p = 0.048, R2 = 0.78; Figure 3.3c). On the Pacific slope, by contrast, there was no significant relationship of elevation preference (or mean elevation of occurrence) with CTmax (Deming regression: p = 0.93, Figure 3.3b; weighted regression: F1,2 = 0.86, p = 0.45, R2 = 0.30; Figure 3.3d). We caution, however, that our analysis on the Pacific slope suffers from very low statistical power (n = 4), so the negative results on the Pacific slope should not be interpreted as a lack of effect of heat tolerance on elevation preference and mean elevation of occurrence.    47     Figure 3. 3: Heat tolerance (CTmax) explains elevation preference and mean elevation of occurrence (MEO) on the Atlantic slope (a and c), but not on the Pacific slope (b and d). Elevation preference was measured as the coefficient of the effect of elevation on abundance. Symbols represent the larvae of the mosquito Culex erethyzonfer, the chironomid Polypedilum spp., the marsh beetle Scirtid spp., the crane fly Trentepohlia spp., and the damselfly M. modesta. Line represents best fit of statistically significant relationship (α = 0.05).     48   Potential effects of species interactions explained the effects of elevation on insect abundance on the Pacific slope, but not on the Atlantic slope (Table 3.2, Figure 3.4). On the Pacific slope, elevation had indirect positive effects on the relative abundances of the larvae of chironomids (Polypedilum spp.) and marsh beetles (Scirtid spp.), and for both taxa this indirect effect was mediated by crane fly larvae (Trentepohlia spp.; pathway ab: Table 3.2b, Figure 3.4e,f), suggesting that chironomids and marsh beetles become more abundant towards higher elevations because of adverse effects of crane flies at low elevations (pathway b: Table 3.2b, Figure 3.4d-f). In contrast to the Pacific slope, species interactions did not mediate effects of elevation on relative abundance patterns of bromeliad insects on the Atlantic slope (pathways ab, ed, and acd: Table 3.2a, Figure 3.4a-c). Instead, the relative abundances of the larvae of mosquitoes (C. erethyzonfer) and marsh beetles were positively affected by elevation, while chironomids and damselflies were negatively affected by elevation (pathways f and a: Table 3.2a, Figure 3.4a-c). Importantly, although crane flies had negative effects on mosquitoes and chironomids on the Atlantic slope (pathway d: Table 3.2a, Figure 3.4a,b), those negative effects of crane fly larvae did not result in any indirect effects of elevation on these insects (pathway ed), presumably because crane flies were not affected by elevation on the Atlantic slope (pathway e: Table 3.2a, Figure 3.4a-c). When we replaced damselfly abundance in analyses with “obligate predator” abundance (i.e., combined abundances of the diving beetle Copelatus spp. and the damselfly), the results were qualitatively identical (Appendix B.3).     49     Table 3. 2: The abundance of each of three focal species (columns) is modelled with structural equation models (SEMs), with each SEM being composed of multiple paths (rows, with letters corresponding to Figure 3.4). The significance (p values) of individual paths is assessed with Z-scores. The SEMs were fit separately for a) the Atlantic slope and b) the Pacific slope. a) Atlantic slope C. erethyzonfer Polypedilum spp. Scirtid spp. path Z p Z p Z p a: M. modesta ~ elevation -2.272 0.023 -2.272 0.023 -2.272 0.023 b: focal species ~ M. modesta -0.280 0.775 1.297 0.195 -1.523 0.128 c: Trentepohlia spp.~ M. modesta -1.472 0.141 -1.472 0.141 -1.472 0.141 d: focal species ~ Trentepohlia spp. -2.227 0.026 -3.404 0.001 -0.612 0.540 e: Trentepohlia spp.~ elevation -0.160 0.873 -0.160 0.873 -0.160 0.873 f: focal species ~ elevation 3.099 0.002 -4.055 < 0.001 3.327 0.001 ab: indirect effect mediated by M. modesta 0.283 0.777 -1.126 0.260 1.265 0.206 ed: indirect effect mediated by Trentepohlia spp. 0.160 0.873 0.160 0.873 0.155 0.877 acd: indirect effect mediated by M. modesta +          Trentepohlia spp. -1.080 0.280 -1.161 0.245 -0.549 0.583  b) Pacific slope C. erethyzonfer Polypedilum spp. Scirtid spp. path Z p Z p Z p a: Trentepohlia spp.~ elevation -4.223 < 0.001 -4.223 < 0.001 -4.223 <0.001 b: focal species ~ Trentepohlia spp. 1.050 0.294 -3.700 < 0.001 -2.693 0.007 c: focal species ~ elevation 1.310 0.190 -1.705 0.088 1.297 0.195 ab: indirect effect mediated by Trentepohlia spp. -1.019 0.308 2.783 0.005 2.270 0.023     50   Figure 3. 4: Path diagram summarizing results of structural equation models for the three focal species on the Atlantic (left panels) and Pacific (right panels) slopes. The three focal species are the larvae of the mosquito Culex erethyzonfer (a and c), the chironomid Polypedilum spp. (b and e), and the marsh beetle Scirtid spp. (c and f). Species that can potentially mediate the effects of elevation on the focal species are the facultative predatory crane fly larva Trentepohlia spp. (both Pacific and Atlantic slopes) and the obligatory predatory damselfly nymph M. modesta (Atlantic slope only). Color indicates statistically significant (black) or non-significant (grey) path (α = 0.05). Thickness of arrow represents magnitude of effect when effect is significant. Direction of effect is indicated by (+) or (-).     51  3.5 DISCUSSION  Species distributions and the make-up of ecological communities are governed by abiotic conditions and biotic interactions, but the interplay between these two factors is still poorly understood. We showed that the relative importance of temperature and species interactions in determining the elevational distributions of bromeliad-dwelling insect communities can depend on environmental conditions, consistent with Prediction 3. On the moist Atlantic slope of Monteverde, insects with high heat tolerances preferred low elevations where conditions are warmest, while insects with low heat tolerances preferred high elevations where it is coldest. On the drier Pacific slope, by contrast, the elevational abundance patterns of insects reflected potential negative interactions from crane fly larvae, which are known to become predatory under drought conditions (Amundrud and Srivastava 2019; Amundrud et al. 2019), consistent with Prediction 3.2. We stress that our results cannot be interpreted in terms of one factor (heat tolerance) being solely responsible for driving elevational distributions on the Atlantic slope, while another (species interactions) is solely responsible for driving the results on the Pacific slope. Instead, both are important, but the relative importance of (and our ability to detect) these factors depended on mountain side and thus environmental context.  Our conclusion from the Atlantic slope, that the impact of physiological heat tolerance on insect elevational distributions overshadows any effects of species interactions, is consistent with expectations from the thermal adaptation hypothesis (Sunday et al. 2012), and supports the findings of other studies that link species’ thermal tolerances with their distributions along gradients in both latitude (Addo-Bediako et al. 2000; Hoffmann et al. 2002) and elevation (Gaston and Chown 1999; Oyen et al. 2016). However, our finding that negative interactions 52  from crane fly larvae outweighed any physiological effects from high temperatures in driving insect elevational distributions on the Pacific slope underlines that, under certain conditions, factors other than physiological traits can be more important in determining species distribution and abundance patterns.  Indirect effects of climate stressors on entire communities can be driven by “leverage species”, that is, species that are both sensitive to the stressor and dominant in terms of species interactions (Harley et al. 2006). Such species can act as “biotic multipliers” of climate change, resulting in community-level effects that outweigh any direct effects of climate change on individual species (Zarnetske et al. 2012). On the Pacific slope, the crane fly Trentepohlia spp. may have acted as leverage species, as it responded both strongly to changes in climatic conditions, while also exerting disproportionate community level effects. The dry and warm conditions on the low elevation Pacific slope resulted in changes in population size as well as potential per capita effects of crane flies, both of which can result in ecosystem level changes (Sanford 1999). Not only were crane fly larvae disproportionately abundant on the low elevation Pacific slope, presumably due to the combined effects of a high heat tolerance  and a high drought tolerance (Amundrud and Srivastava 2015), but the warm and dry conditions also facilitated negative effects of crane flies on other bromeliad insects, consistent with recent findings that crane fly larvae become ferocious predators of other insect larvae under drought (Amundrud and Srivastava 2019; Amundrud et al. 2019). The mosquito C. erethyzonfer was the only potential prey that was not affected by crane flies on the Pacific slope (Table 3.2b, Figure 3.4d). Instead, the preference of this mosquito for high elevations was most likely linked to its low tolerance to heat. On the Atlantic slope, by contrast, no insects acted as leverage species to transmit the effects of altered climatic conditions on the community. Instead, the taxon that 53  exerted adverse effects on other insects (crane fly) was not affected by elevation, while the taxon that strongly responded to elevation (damselfly) was not common enough to exert significant effects on the community (Table 3.2a, Figure 3.4a-c). Collectively, these results suggest that under certain environmental conditions, indirect effects mediated by species interactions may override any direct physiological effects of environmental conditions on species distributions. Pinpointing these conditions is critical when making predictions about species distributions, particularly in the face of climate change.  Making accurate predictions of how climate change will affect species distributions and communities is ever more important in a rapidly changing world. While species’ physiological tolerance limits to abiotic conditions have often been employed to predict ecological responses to climate change, they may not always be the main drivers of species distributions. Not only is it becoming increasingly clear that single species can act as biotic multipliers of climate change (Zarnetske et al. 2012), and that such biotic multipliers can be more important drivers of species distributions than species’ physiological limits to abiotic conditions (Ockendon et al. 2014), but we show here that the relative importance of abiotic and biotic effects can depend on environmental context. It is thus likely that the relative importance of physiological and ecological drivers of species distributions will shift with changing abiotic conditions, such as climate change. In our study, insect elevational distributions on the Atlantic slope were related to their physiological tolerance limits to high temperatures. However, with climate change, the effects of species interactions (i.e., from predation) may override the effects of physiological limits to drive insect distribution patterns. Specifically, predatory damselflies are expected to benefit from warmer temperatures and increase locally in abundance. In addition, this predator will likely track climate change to higher elevations, as previous species distribution models 54  suggest that this species’ range is limited by low temperatures found at high elevations (Amundrud et al. 2018). As such, increased predation pressure from this keystone predator on the Atlantic slope will potentially alter insect elevational distributions as the climate warms. Altered top-down effects from climate change are predicted to affect food webs across many ecosystem types (Marino et al. 2018), underlining that predators are particularly likely to mediate the effects of climate change (Urban et al. 2017).   We acknowledge several caveats of our study. First, the decrease in bromeliad size towards higher elevations could have potentially influenced elevational abundance patterns of bromeliad insects. However, we can discount this possibility. The abundances of only two insect taxa were related to bromeliad size: both Scirtid spp. and Polypedilum spp. increased with bromeliad size. If the elevational abundances of these insects were driven by bromeliad size, we would expect a decrease in their abundances with elevation, along with the decrease in bromeliad size. This was not the case, however, as Scirtid spp. abundance increased towards higher elevations, while Polypedilum spp. abundance was not affected by elevation. Second, we recognize that there are several obstacles to experimentally determining physiological tolerance limits of insects. We minimized any potential biases that could have affected the results by applying the same strict experimental protocol to all subjects (e.g., constant rates of temperature ramping and close to constant starting temperatures, identical selection of endpoints to estimate heat tolerance, as well as acclimation times of at least 4 days). We stress, however, that while our heat tolerance estimates are comparable within this study, they should not be compared to other studies where the methodology might have differed (Lutterschmidt and Hutchison 1997; Terblanche et al. 2007; Cooper et al. 2008). Third, we acknowledge that, although multiple studies have suggested that the most important predictor of species distributions is their 55  physiological tolerance to extreme heat events (e.g., Kellermann et al. 2012; Vorhees et al. 2012; Overgaard et al. 2014), tolerance to low temperature may be more important in governing geographic distributions for some species (Lutterschmidt and Hutchison 1997). M. modesta damselfly larvae in particular were likely limited by cold temperatures towards higher elevations, based on the results from a species distribution model (Amundrud et al. 2018). Fourth, we showed that the relative importance of physiological tolerance limits and species interactions depended on drought. However, it is likely that the dynamic nature of the relative importance of these two factors also shifts with season, and it would be informative to repeat this study across seasons and years. Finally, our finding that damselflies were very rare at the lowest elevation site on the Atlantic slope, even though this species tended to become more common in other lower elevation sites on the Atlantic slope, was surprising. The low damselfly numbers at this site may be related to the high abundances of diving beetles (Figure 3.1), as M. modesta damselflies are known to avoid these predators due to negative effects of intra-guild predation or tactile stimulation (Atwood et al. 2014).   We studied the interplay of thermal tolerance limits and species interactions in shaping aquatic insect food webs inside Neotropical bromeliads along two elevational gradients, and demonstrated that the relative importance of abiotic and biotic drivers of insect distributions can depend on environmental context. While heat tolerance was more important in explaining insect elevational distributions when conditions were wet (Atlantic slope), indirect effects from potential predators were more important in explaining their elevational distributions when conditions were dry (Pacific slope). These two elevation gradients allowed us to examine species occurrence over a broad range of climate parameter space but over a spatial extent small enough that dispersal is not expected to be limiting, enabling us to attribute distributions to the 56  interactive effects of rainfall, temperature and potential predators. The mechanistic processes inferred from this landscape-scale study may inform our understanding of regional and global-scale patterns in community structure where dispersal limitation, phylogeographic constraints and covariance between biotic and abiotic variables can make causal inference weaker (Araújo and Luoto 2007). Complex community-level responses to climate change are still poorly understood, and these findings support other recent studies demonstrating that the effects of climate change on species ranges and community dynamics may be highly influenced by species interactions (Clark et al. 2011; Diamond et al. 2016). Importantly, our study also demonstrates the dynamic nature of the interplay between abiotic and biotic selective pressures on insect physiology and distributions. This is an important insight given the rapid rates of changing environmental conditions globally. As the climate changes, tropical montane regions, which often contain biodiversity hotspots with exceptionally high rates of endemism, are of particular concern (Graham et al. 2014). In all ecosystem types, however, the relative importance of abiotic and biotic effects on species distributions is likely to shift with climate change. Future research should thus focus on understanding how climate change alters the factors that mediate species distributions.      57  CHAPTER 4   4 DROUGHT ALTERS THE TROPHIC ROLE OF AN OPPORTUNISTIC GENERALIST IN AN AQUATIC ECOSYSTEM3                                                                 3 Previously published as Amundrud, S. L., Clay-Smith, S. A., Flynn, B. L., Higgins, K. E., Reich, M. S., Wiens, R. H. W., and Srivastava, D. S. (2019) Drought alters the trophic role of an opportunistic generalist in an aquatic ecosystem. Oecologia 189: 733-744  58  4.1 CHAPTER SUMMARY  Abiotic change can alter species interactions by modifying species’ trophic roles, but this has not been well studied. Until now, bromeliad-dwelling tipulid larvae were thought to positively affect other macroinvertebrates via a facilitative processing chain. However, under drought, we found the opposite. We performed two microcosm experiments in which we factorially manipulated water level and predation by tipulids, and measured the effects on mosquito and chironomid larvae. The experiments differed in whether high water was contrasted with low or no water, allowing us to distinguish between the effects of desiccation stress (no water) and increased encounter rates due to compression of habitat or reductions in prey mobility (low and no water). We also included a caged tipulid treatment to measure any non-consumptive effects. As well as directly reducing prey survival, reductions in water level indirectly decreased chironomid and mosquito survival by altering the trophic role of tipulids. Our results suggest that increased encounter rates with prey led to tipulids becoming predatory under simulated drought, as tipulids consumed prey under both low and no water. When water level was high, tipulids exerted negative non-consumptive effects on prey survival. Because opportunistic predators are common throughout aquatic ecosystems, the effects of drought on the trophic roles of species may be widespread. Such restructuring of food webs should be considered when attempting to predict the ecological effects of environmental change.      59  4.2 INTRODUCTION   Many ecosystems are characterized by fluctuating environments, and climate change threatens to affect ecosystem structure and functioning by altering patterns in environmental heterogeneity (Boyd et al. 2016; Vázquez et al. 2017). Abiotic environmental variation, such as seasonal fluctuations or spatial variation in temperature and precipitation, can structure food webs through direct physiological effects on species, as well as indirectly by altering species interactions (McMeans et al. 2015; Humphries et al. 2017). Species interactions can respond to environmental variation when abiotic conditions determine the co-occurrence or the functional responses of species (Humphries et al. 2017). The relative importance of trophic control, for instance, is often governed by environmental heterogeneity, as environmental conditions can affect consumer impacts within a food web (Power 1992). By affecting species interactions, global climate change has the potential to alter food web composition and ecological processes (McMeans et al. 2015). However, we still know little about how the effects of climate change on environmental heterogeneity alter biotic processes, particularly by affecting trophic interactions (Suttle et al. 2007; Tylianakis et al. 2008; Traill et al. 2010; Blois et al. 2013).   Environmental heterogeneity can influence trophic control when consumers exploit different trophic levels in different habitats (Tunney et al. 2012) or during different seasons (Carnicer et al. 2008). Variable environments can affect trophic interactions by altering the relative abundances (Ancona et al. 2012; Amundrud and Srivastava 2016) or the spatiotemporal overlap (Winder and Schindler 2004b; Wolkovich et al. 2012) of consumers and resources, as well as by altering the per capita interaction strengths of consumers (Kruse et al. 2008; O’Connor 2009; McCluney and Sabo 2009). For instance, flexible consumers like omnivores, who can feed 60  on multiple trophic levels (Pimm and Lawton 1978; Kratina et al. 2012) and shift their diets based on the availability or profitability of resources (Holt and Polis 1997), frequently shift their trophic role in accordance with changing environmental conditions (Gestoso et al. 2015; Törnroos et al. 2015; MacAvoy et al. 2016). Such diet shifts may occur if environmental variation affects consumer-resource encounter rates (e.g., Boukal 2014) or the metabolic needs of consumers (e.g., McCluney and Sabo 2009).   First, environmental heterogeneity can determine the trophic structure of an ecosystem if environmental conditions affect consumer-resource encounter rates (McMeans et al. 2015). For example, habitat compression in drying water bodies can lead to increased predator-prey encounter rates through increased prey densities (Dudgeon 1993; Magoulick and Kobza 2003; McHugh et al. 2015). Consumer-resource encounter rates can also be altered when environmental change alters predator foraging traits and/or prey vulnerability traits. Warming, for instance, can lead to altered predator-prey interactions through asymmetric effects on their respective body velocities (Dell et al. 2014). For example, increases in predatory damselfly activity levels due to warmer temperatures interact with changes in their body sizes to increase cannibalism and thus shift trophic interactions (Start et al. 2017). Prey vulnerability traits can also depend on environmental conditions. For instance, elevated CO2 has been shown to decrease caterpillar body size and increase their vulnerability to predation (Coll and Hughes 2008). Abiotic conditions can also affect the ability of prey to detect predators (van Uitregt et al. 2013; Marino et al. 2017) or alter their propensity to flee from predators (de Barros et al. 2010).   Second, environmental heterogeneity can regulate the trophic role of consumers when new metabolic demands necessitate a shift in diet (Rall et al. 2010; Vucic-Pestic et al. 2011; McCluney 2017). For example, in a controlled laboratory experiment, warming resulted in a shift 61  in dietary preference of grasshoppers from mainly carbohydrates to increased protein (Schmitz et al. 2016). Similarly, predatory spiders consume crickets when stressed by desiccation, but not when water is available (McCluney and Sabo 2009). Although temperature has been shown to alter the functional and/or trophic roles of consumers (Petchey et al. 2010; Rall et al. 2010; Sentis et al. 2014; Lemoine 2017), far less studies have considered the possibility of altered precipitation regimes affecting trophic links in food webs through changes in physiological demands.  Here, we studied the effects of drought, an important stressor that is predicted to increase in both frequency and severity with climate change in some areas of the world (Prudhomme et al. 2014), on the trophic role of tipulid (Tipulidae: Diptera) larvae found in a Neotropical aquatic food web. Tank bromeliads (Vriesea gladioliflora H. Wendl, Vriesea sanguinolenta Cogn. and Marchal, and Guzmania scherzeriana Mez) are small detrital-based ecosystems that contain aquatic food webs consisting of macroinvertebrates and microorganisms (Srivastava and Bell 2009). In this ecosystem, drought is an important stressor that determines both species occurrences (Amundrud and Srivastava 2015) and species interactions (Amundrud and Srivastava 2016; Pires et al. 2016; Marino et al. 2017). Until now, bromeliad-dwelling tipulid larvae were considered facultative detritivores (Merritt and Cummins 1995) and thought to facilitate other bromeliad macroinvertebrates through conditioning litter (Starzomski et al. 2010). However, occasional opportunistic predatory behaviour under drought conditions has been observed (Srivastava, Amundrud pers. obs.; Amundrud & Srivastava, 2016), leading us to hypothesize that drought may alter the trophic role of bromeliad tipulids.  We performed two microcosm experiments in which we examined individual and combined effects of tipulid predation and water level on two dominant groups of bromeliad 62  insects, the larvae of mosquitoes and chironomids. Experiment 1 contrasted tipulid predation in ambient water level conditions and in the complete absence of water. Experiment 2 contrasted tipulid predation in ambient water level conditions and at reduced water levels. If tipulids become predatory because reductions in the water level increase encounter rates with prey (for instance by concentrating prey in space or reducing prey mobility), we hypothesize that the absence of water (Experiment 1) and a low water level (Experiment 2) should both lead to predation by tipulids. However, if tipulids become predatory because desiccation stress alters their metabolic needs, we hypothesize that only the complete absence of water (Experiment 1) should lead to predation by tipulids, as the low water level in Experiment 2 was designed to avoid any desiccation stress. In addition, we considered whether effects of tipulids on the insect community were purely consumptive, or also non-consumptive, by contrasting the effect of caged, uncaged, and no tipulids (Experiment 1).   4.3 METHODS 4.3.1 Study Site and System  To test for any interactive effects of tipulid predation and water level, we performed two microcosm experiments in which we factorially manipulated drought and tipulid predation, and measured the effects on artificially assembled communities consisting of mosquito and chironomid larvae. Experiment 1 was carried out in the fall of 2015 in a laboratory in Monteverde (10°18'N 84°48'W), Costa Rica. Experiment 2 was carried out in the fall of 2016 on 63  a shaded outdoor veranda at the nearby University of Georgia Costa Rica (UGA-CR) campus (10°16'N 84°47'W). In this part of Costa Rica, climate change is predicted to significantly reduce cloud cover and rainfall by up to 30% (Karmalkar et al. 2008), thereby increasing the frequency and severity of drought. The macroinvertebrate food web contained in tank bromeliads consists predominantly of the aquatic larvae of dipterans such as detritivorous crane flies (Trentepohlia spp., in Tipulidae) and chironomids (e.g., Polypedilum spp. and Orthocladiinae), and filter-feeding mosquitoes (e.g., Anopheles spp., Culex spp., and Wyeomyia spp.), as well as detritivorous aquatic beetle larvae (Scirtes spp.,Starzomski et al. 2010).   4.3.2 Experimental manipulation of tipulid predation and water level  Microcosms, which mimic a typical bromeliad leaf  well (Amundrud and Srivastava 2015, 2016), were necessary to precisely manipulate drought and community composition. Microcosms consisted of 50 mL centrifuge tubes rendered semi-permeable by cutting off the tapered bottoms and part of the side and covering these openings with 80 µm mesh. The tubes were placed mesh-bottom down in plastic cups so that, by altering the water level in the cups, we could manipulate the water level in the tubes without disturbing the contained communities. To each microcosm tube, we added oven-dried detritus that had been rehydrated for 24 hours in the form of a single leaf of Conostegia xalapensis Bonpl. (Melastomataceae) and particulate organic matter (80–850 μm diameter) obtained from bromeliads. To re-establish microbial communities on detritus that may have been lost through drying, we also added 2 mL of filtered (through 80 µm mesh) water obtained from natural bromeliads as a microbial inoculant. We monitored insect emergences with emergence traps attached to the tops of each microcosm.   64  4.3.2.1 Experiment 1: Effects of tipulids in the presence and absence of water  We factorially manipulated water (2 levels: presence and absence of water) and tipulids (3 levels: no tipulid, caged tipulid, and uncaged tipulid). Bromeliad insects in Costa Rica are able to survive in the complete absence of water for at least 18 days (Amundrud and Srivastava 2015), and species interactions in this system have been shown to be affected by drought lengths of up to 22 days (Amundrud and Srivastava 2016). As any negative effects of tipulids on the prey community could result from either consumptive or non-consumptive effects, we controlled for any non-consumptive effects (such as inducing a fear response in potential prey or by affecting water quality and microbial densities) by also including a caged tipulid treatment.   We acquired tipulid larvae, as well as the most common species of chironomid and mosquito larvae, from bromeliads at sites up to 10 km from the village of Santa Elena at Monteverde. We assembled 54 prey communities (n = 10 of each tipulid level for the water present treatment; n = 8 of each tipulid level for the water absent treatment) in separate microcosms. Each prey community consisted of 10 mosquitoes [5 Wyeomyia (subgenus Hystatomyia) sp. and 5 Culex (subgenus Micraedes, either Culex jenningsi or Culex erethyzonfer)] and 10 chironomids (5 Polypedilum spp., and 5 Orthocladiinae spp.). We initiated the experiment on 21-Nov-2015 (water present treatment) and 2-Dec-2015 (water absent treatment). Because we were not able to find the same species of Culex for the second set-up, the identity of the Culex species was confounded with water treatment (all water present treatments received C. jenningsi, all water absent treatments received C. erethyzonfer). However, we show below that any effects of water treatment in our experiment are not driven by the confounding of Culex species identity with water treatment (see Results).  65   We randomly assigned each community to one of three tipulid treatments (no tipulid; caged tipulid; and uncaged tipulid). We constructed tipulid cages using mesh bags (dimensions: 3 cm x 1 cm; mesh size: 150 µm), and caged tipulids received half of the leaf inside their cage as a food source, which sustained them to the end of the experiment (i.e., no leaf was completely consumed). Communities that were assigned to the no-tipulid or uncaged tipulid treatments received an empty mesh bag (containing neither a tipulid nor a leaf) to control for any cage effects.   To quantify any treatment effects on tipulid growth rate and control for differences in tipulid sizes between treatments, we measured body lengths of tipulids before and after the experiment. We used ImageJ (Schneider et al. 2012) to measure body lengths from digital photographs taken through a dissecting microscope. Again, we were careful to choose tipulids that were as similar as possible in their initial size for our experiment, and then confirmed that tipulid sizes were randomly distributed among treatments (Tipulid treatment: F1,21 = 0.47, P = 0.50; Water treatment: F1,21 = 1.32, P = 0.26; interaction: F1,21 = 0.03, P = 0.86).   We filled microcosm cups with filtered stream water to reach a level of 15 mL (i.e., a water depth of roughly 2.1 cm) inside the microcosm tubes to represent typical water levels in a bromeliad leaf well (Amundrud and Srivastava 2015, 2016), and placed microcosms indoors to ensure that the temperature was constant between the two experimental runs. After 24 hours of insect acclimation to the microcosms, we removed all water from microcosms assigned to the water absent treatment. This drying rate is identical to that used in similar experiments studying the effects of drought on bromeliad communities (Amundrud and Srivastava 2015, 2016), and field observations show that Costa Rican bromeliads can lose as much as 1-3 cm of water depth a day (Diane Srivastava, unpublished data). The aim of the water absent treatment in our 66  experiment was to test for sublethal effects of drought on species interactions. Thus, we chose a drought length of nine days, a length that has previously been shown to not significantly affect bromeliad insect survival in environments with high humidity (Amundrud and Srivastava 2015, 2016), while still allowing enough time for communities to respond to the treatments. During the experiment, we changed 50% of the water in microcosms subjected to the water present treatment every three days to mimic the natural flushing of bromeliads by rain. We checked and recorded emergences daily.   Nine days after initiation we added water to the microcosm cups subjected to the water absent treatment until 15 mL inside the microcosm tubes were reached. Twenty-four hours later, we emptied microcosm contents in random order and searched for surviving larvae and pupae. Before analysis, we excluded the few replicates where tipulids died or escaped, resulting in final sample sizes of five to eight microcosms for each treatment.   4.3.2.2 Experiment 2: Tipulid predation in high vs. low water  In this experiment we shift our focus from the effects of water presence vs. absence to a partial reduction in water level, in order to isolate the effects of habitat compression (imposed by both the absence of water and low water level) from those of desiccation stress (imposed only by the absence of water). We factorially manipulated water level (2 levels: high and low) and tipulid state (2 levels: alive or dead). Note that the high water level treatment is identical to the water present treatment in Experiment 1. We had initially planned a stressed tipulid treatment, but all tipulids died as a result of the pre-experimental drought stress, likely because of faster drying rates at the warmer and dryer lower elevation site of Experiment 2. We have kept the dead tipulid 67  treatment in the experiment as it unexpectedly provided a predation-free control. Although dead tipulids could have influenced prey through decomposition products, we will show in the results that this did not happen over the timescale of the experiment. Again, we acquired tipulid, chironomid, and mosquito larvae from bromeliads at sites up to 5 km from Monteverde, and assembled prey communities consisting of 4 chironomids (2 Polypedilum spp., and 2 Orthocladiinae spp.) and 4 mosquitoes [2 Wyeomyia (subgenus Hystatomyia) sp. and 2 Culex jenningsi].  Because of the difficulty of acquiring enough insect larvae at the same time, we divided the experiment into 3 temporal blocks which were initiated on 20-Oct-2016 (9 microcosms), 2-Nov-2016 (26 microcosms), and 17-Nov-2016 (18 microcosms). We placed microcosms outdoors under cover to provide shade and prevent natural inputs of precipitation. After assembling communities in microcosms, we reduced the water in the low water level treatments to 3 mL (i.e., 20% of the high water level) inside the microcosm tubes by removing water from the cups. 24 hours after set-up, we searched for surviving larvae. We then returned the microcosm contents to the tubes, and repeated this procedure for up to five days. Before analysis, we excluded the few “live tipulid” replicates in which tipulids subsequently died, disappeared or escaped, resulting in final sample sizes from 12 to 16.     68  4.3.3 Statistical analysis 4.3.3.1 Experiment 1: Effects of tipulids in the presence and absence of water  We quantified the effects of water, tipulids, and their interaction on total remaining prey abundance (i.e., the surviving larvae, pupae, and emerged adults of mosquitoes and chironomids together), as well as separated by taxonomic family (i.e., the surviving larvae, pupae, and emerged adults of mosquitoes and chironomids separately), using linear models. To disentangle the effects of predation from any non-consumptive effects of tipulids, we employed t-tests to explore non-consumptive (no vs. caged tipulid) and consumptive (caged vs. free tipulid) effects on prey as post hoc analyses when tipulid treatment, or its interaction with water treatment, was significant. Because the identity of the Culex species (Culex jenningsi vs. Culex erethyzonfer) was confounded with water treatment, we repeated the above analysis on the data from which all Culex spp. were removed from the total prey count. Although this second analysis does not rule out the possibility that Culex spp. had differential effects on other species, it enabled us to show that any effects of water treatment were not driven by the confounding of Culex species identity and water level. We used a linear model to determine the individual and interactive effects of water treatment and cage treatment on tipulid growth rates (as relative changes in body length for surviving tipulids).   For all models, our data met model assumptions (as determined by Quantile-Quantile, residual and outlier plots) without requiring transformation of the response or predictor variables. We carried out all analyses in the programming language R (version 3.2.1: R Core Team, 2015). The data to Experiment 1 are available on KNB Digital Repository (Amundrud et al. 2019).  69  4.3.3.2 Experiment 2: Tipulid predation in high vs. low water  We quantified the effects of water level, tipulids, and their interaction on total prey survival, as well as on chironomid and mosquito survival separately, by employing a mixed effects Cox proportional hazards survival analysis using the R package coxme (Terry M. Therneau 2015). Fixed effects were tipulid treatment, water level, and their interaction. Random effects were microcosm effects nested within temporal block. We obtained P-values by employing an analysis of deviance on the Cox models. The data met model assumptions of proportional hazards, as determined by visual inspection and tests based on the Schoenfeld residuals (Schoenfeld 1980). The data to Experiment 2 are available on KNB Digital Repository (Amundrud et al. 2019).   4.4 RESULTS 4.4.1 Experiment 1: Effects of tipulids in the presence and absence of water  The effects of tipulids on total prey survival depended on water treatment (Table 4.1a, Figure 4.1a). As predicted, there was no prey consumption by tipulids when water was present, but tipulids consumed prey when water was absent, as uncaged tipulids decreased prey abundance by 79 % compared to caged tipulids (Table 4.1b, Figure 4.1b). In contrast, non-consumptive effects of tipulids were only important in the presence of water, where caged tipulids decreased prey abundance by 25 % relative to treatments with no tipulids (Table 4.1b, Figure 4.1c).    70    Table 4. 1: Effects of tipulids and water on mosquitoes and chironomids in Experiment 1. (a) Effect tests based on 2-way ANOVA, as summarized by F ratios and P-values. “Tipulid” represents three levels of tipulid treatments (no tipulid, caged tipulid, and uncaged tipulid), “Water” represents two water treatments (present and absent), and “Tipulid x Water” represents their interaction. Significant P-values after sequential Bonferroni corrections (α1 = 0.05, α2 = 0.025, and α3 = 0.017, α’s are sequentially compared from largest to smallest P -value in each column) are indicated in bold. (b) Partitioning of tipulid effects into consumptive and non-consumptive effects. “Consumptive” effects compare caged vs. uncaged tipulid treatments for “Water present” and “Water absent” treatments, respectively. “Non-consumptive” effects compare no vs. caged tipulid treatments for “Water present” and “Water absent” treatments, respectively. Significant P-values after sequential Bonferroni corrections (α1 = 0.025, α2 = 0.013, and α3 = 0.008; α’s are sequentially compared from largest to smallest P -value in each column) are indicated in bold. Because of non-orthogonality, α’s are halved.  (a)  2-way ANOVA Tipulid Water Tipulid x Water Response F2,36 P F1,36 P F2,36 P Total prey abundance 12 < 0.001 151 < 0.001 12 < 0.001 Chironomid abundance 10 < 0.001 6.6 0.014 7.7 0.002 Mosquito abundance 2.6 0.092 261 < 0.001 3.6 0.037  (b)  Tipulid effects Consumptive Non-consumptive  Water present Water absent Water present Water absent Response t13 P t9 P t13 P t11 P Total prey abundance -1.1 0.289 6.7 < 0.001 4.0 0.001 -0.47 0.64 Chironomid abundance -1.0 0.322 5.5 < 0.001 3.2 0.007 0.33 0.75 Mosquito abundance -0.14 0.891 3.9 0.004 3.6 0.003 -0.92 0.38    71  Figure 4. 1: Left Panel: Effects of water and tipulids on (a) total prey abundance (chironomids and mosquitoes combined), (d) chironomid abundance, and (g) mosquito abundance in Experiment 1. Middle Panel: Impact of tipulid predation in the presence and absence of water. (b) Number of total prey consumed (chironomids and mosquitoes combined), (e) number of chironomids consumed, and (h) number of mosquitoes consumed. Number of consumptive deaths is calculated as the difference of prey abundance in communities with caged vs. uncaged tipulids. Error bars represent 95% confidence intervals. Right Panel: Impact of non-consumptive effects of tipulids in the presence and absence of water. (c) Number of total prey deaths (chironomids and mosquitoes combined), (f) number of chironomid deaths, and (i) number of mosquito deaths. Number of non-consumptive deaths is calculated as the difference of prey abundance in communities with no vs. caged tipulids. Error bars represent 95% confidence intervals. Chironomids are Polypedilum spp. and Orthocladiinae spp., mosquitoes are Wyeomyia spp. and Culex spp.  72   To further understand the responses of prey, we examined chironomid and mosquito survival separately (prey emergence rates during the experiment were low, averaging less than 4 % per microcosm). For both taxa, the effects of tipulids on the total number of surviving prey again depended on water treatment (Table 4.1a). Consistent with the results for total prey, consumptive effects of tipulids significantly decreased the survival of both chironomids and mosquitoes (by 71 % and 100 %, respectively) when water was absent, while consumption by tipulids was not important when water was present (Table 4.1b, Figure 4.1e,h). Non-consumptive effects, on the other hand, arose only in the presence of water, where they decreased chironomid and mosquito abundance by 37 % and 14 %, respectively (Table 4.1b, Figure 4.1f,i). While the removal of water by itself had a weak direct effect on chironomids, whose abundances decreased by only 20 % (Figure 4.1d), it decreased mosquito abundance by 83% (Figure 4.1g).  When we repeated the analyses on the subset of the data excluding Culex spp., the results for total prey abundance were qualitatively similar, signifying that the above results were not driven by the confounding of Culex spp. identity with water level (Appendix C.1). Although tipulids were able to consume prey in the absence of water, this cage x water level interaction did not affect the growth rates of tipulids (F1,17 = 0.09, P = 0.77; Appendix C.2).  4.4.2 Experiment 2: Tipulid predation in high vs. low water  As expected, the effects of tipulids on total prey survival depended on water level (Table 4.2). When tipulids were alive, prey experienced high mortality (59 %) only at low water level, compared to a mortality of 30 % at high water level (Figure 4.2a). The latter is comparable to prey mortality when tipulids were dead (23 % for high and 21 % for low water level). 73      Table 4. 2: Effects of water level and tipulids on chironomid and mosquito survival in Experiment 2. “Tipulid” represents two levels of tipulid treatments (alive and dead), “Water Level” represents two levels of water level treatment (high and low), and “Tipulid x Water Level” represents their interaction. Significant P -values after sequential Bonferroni corrections (α1 = 0.05, α2 = 0.025, and α3 = 0.017; α’s are sequentially compared from largest to smallest P -value in each column) are indicated in bold.  Experiment 2   Tipulid Water Level Tipulid x Water Level Response Χ2 P Χ2 P Χ2 P Total prey survival 38 < 0.001 5.6 0.018 6.2 0.013 Chironomid survival 32 < 0.001 6.2 0.013 0.99 0.32 Mosquito survival 16 < 0.001 1.8 0.18 9.7 0.002    74   Figure 4. 2: Survivorship (left panel) and number of surviving larvae for each point in time (right panel) from Experiment 2. (a) and (b) show total prey (chironomids and mosquitoes combined), (c) and (d) show chironomids, and (e) and (f) show mosquitoes    75   To further understand prey responses, we then examined chironomid and mosquito survival separately. Interestingly, chironomid survival was not affected by the interaction of tipulid treatment with water level (Table 4.2). Instead, chironomid survival was significantly reduced independently by live tipulids (62 % mortality compared to 29 % mortality when tipulids were dead) and low water level (48 % mortality compared to 40 % mortality when water level was high; Figure 4.2b). For mosquito survival, on the other hand, negative effects of tipulid predation did depend on low water level (Table 4.2). When tipulids were alive, mosquitoes experienced a relatively high mortality (48 %) when the water level was low, compared to when water level was high (17 %; Figure 4.2c). The latter is comparable to mosquito mortality for treatments in which tipulids were dead (21 % for high and 11 % for low water level).    4.5 DISCUSSION   Altered environmental conditions can affect animal communities by altering species interactions. Here, we provide mechanistic insights into how drought stress affects a microecosystem through altering trophic links. In both experiments, simulated drought altered the trophic role of tipulids, who became opportunistic predators when the water column was reduced or removed. While tipulids had negative non-consumptive effects on prey at high water levels, consumptive effects of tipulids were restricted to treatments where water was absent or reduced.  There are several potential mechanisms that can explain how drought enabled tipulids to consume animal prey. First, drought could have caused mosquitoes and chironomids to become 76  vulnerable to predation by negatively affecting their locomotor performance or behavioural defence mechanisms. Tipulids employ crawling as the primary locomotive mode for foraging inside and outside the water, while swimming is non-directional and primarily used as evasive action (Pritchard 1985). The locomotive behaviour of chironomid larvae is similar. Chironomids primarily crawl along the substrate, although they can swim through the open water column in short non-directional bursts when disturbed and to escape predation (Brackenbury 2000). The locomotor behaviour of mosquitoes, on the other hand, is fundamentally different in that they can freely swim or glide through the water column (Christophers 1960; Brackenbury 2001). Tipulids stay mobile in the complete absence of water (Pritchard 1985; Experiment 1; Amundrud, personal observation), and although chironomids are still able to move along the substrate, they cannot escape predation from tipulids through rapidly swimming through the water column. For mosquitoes, which have no means of moving along a dry substrate, the absence of water can completely hamper their ability to escape, and indeed, no mosquito survived in treatments with both no water and uncaged tipulids (Experiment 1). In addition to affecting the ability of mosquito and chironomid prey to escape predation, drought could have also affected other behavioural defence mechanisms. Polypedilum chironomids build algal tubes as protection from predators (Macchiusi and Baker 1992; Baker and Ball 1995). By affecting the algal building material through drying, the absence of water could potentially prevent chironomids from building such tubes. Although adverse effects on locomotor ability and prey defence mechanisms from drought present reasonable explanations for predatory effects in the complete absence of water (Experiment 1), they do not fully explain why tipulids also became predatory when the water level was merely reduced (Experiment 2).  77   Second, drought could have enabled tipulids to consume animal prey by compressing the community in space, and thus increasing random encounters of tipulids with prey (Dudgeon 1993; Magoulick and Kobza 2003; McHugh et al. 2015). Inside bromeliad wells, tipulid larvae forage on a two-dimensional surface, whereas their mosquito and chironomid prey can occupy the water column. As a result, encounter rates and subsequent prey capture rates for high water levels are likely low, consistent with our finding that tipulids do not consume mosquitoes or chironomids when communities are subjected to high water levels. For organisms normally suspended in the water column, the complete absence of water (Experiment 1) effectively alters the habitat from a three-dimensional space (i.e., the water column) to a two-dimensional area (i.e., the substrate). While the reduction in water level (Experiment 2) would not change the dimensionality of the habitat, it would still result in habitat compression (Magoulick and Kobza 2003; McHugh et al. 2015). This compression of the community could effectively increase random encounter rates of tipulids with prey, potentially explaining the rise in prey consumption by this opportunistic consumer (McGill and Mittelbach 2006; Ledger et al. 2013a).  While the first two mechanisms are similar in that drought enables tipulids to consume animal prey by increasing predator-prey encounter rates, a third mechanism that could explain how tipulids became predatory under drought is that desiccation stress altered the physiological demands of tipulids, and thus their dietary protein requirements (Schmitz et al. 2016). As food selection by aquatic insect detritivores is often predominantly influenced by their nutritional requirements (Cargill et al. 1985), tipulids stressed by drought may have consumed insect prey in order to meet their altered dietary needs. We think this mechanism is unlikely, as tipulids became predatory both under the complete absence of water (Experiment 1) and under a reduction in water level (Experiment 2), but only the former should have imposed a desiccation stress. It is 78  important to note, however, that even a reduction in water level could have induced a stress response (for example due to increased density and cues of organisms), so we cannot fully reject this mechanism. One may also speculate that tipulids started consuming animal prey because desiccated detrital matter became unpalatable to tipulids in the absence of water, forcing them to eat other larvae. We reject this possibility as a mechanism, as tipulids consumed animal prey in both the absence of water (Experiment 1), as well as the low water level treatment (Experiment 2), but detritus could only dry in the former treatment.   Non-consumptive effects of tipulids on prey survival were also important, but only in the presence of water (Experiment 1). If predator cues are transmitted through the water column, predation risk may have resulted in reductions in the prey’s foraging effort and energy income (Preisser et al. 2005), resulting in reduced prey survival. Alternatively, the observed decreases in prey survival may be a result of adverse effects of tipulids on water quality through enhancing microbial densities. Indeed, visual examination of water color showed that water quickly became cloudy when tipulids were present (live or dead), but not when they were absent (Experiment 1: no tipulid treatment). Increased turbidity with tipulid presence has also been observed in similar experiments conducted at another site in northwestern Costa Rica (e.g., Amundrud and Srivastava 2016).  Over the short duration of these experiments, we were not able to establish that there was a fitness advantage for bromeliad tipulids when they consumed animal prey, as tipulids maintained equivalent growth rates across all treatments. Bromeliad-dwelling tipulids have previously not been considered to be predatory (but see Amundrud and Srivastava 2016), although there is evidence that tipulids in other systems are predatory (e.g., streams: Martin et al., 1980; Woodward & Hildrew, 2001; Panamanian tree holes: Yanoviak, 1999). On the 79  contrary, bromeliad-dwelling tipulids were thought to mainly have positive effects on other bromeliad macroinvertebrates via facilitation by a processing chain, as tipulid shredders have been shown to increase chironomid growth and survival by conditioning litter (Starzomski et al. 2010). The lack of studies documenting negative effects of bromeliad tipulids on other insects likely results from the fact that this species does not consume other animals when communities are submersed in water, that is, the condition under which these systems have been primarily studied.  Fresh waters are particularly vulnerable to climate change, and in several areas of the world the prevalence or duration of drought events in seasonal freshwater ecosystems is likely to increase (Prudhomme et al. 2014). Our study supports the growing amount of evidence that drought events will have major consequences to the structure and functioning of freshwater ecosystems (e.g., Chase and Knight 2003; Ledger et al. 2012, 2013b, a; Amundrud and Srivastava 2016). Drought has been shown to alter the trophic structure in various aquatic ecosystems (e.g., ponds: Chase and Knight 2003; streams: Ledger et al. 2013b; bromeliads: Amundrud and Srivastava 2016); however, in many of these studies drought generally led to a weakening of top-down control by disproportionately affecting species at the top of the food chain (e.g., Chase and Knight 2003; Ledger et al. 2011, 2013b; Amundrud and Srivastava 2016). Our study reveals that drought can also alter the trophic structure of a food web by increasing top-down control when a consumer has access to prey that were inaccessible under no-drought conditions. In a system where precipitation is expected to decrease (Karmalkar et al. 2008) and drought is expected to increase (Zotz and Thomas 1999), this has important implications for community structure and ecosystem functioning. The negative indirect effects of drought on mosquitoes and chironomids through increased top-down control will act in synergy to any direct 80  physiological effects of drought on these species. We thus expect a shift in species composition away from species vulnerable to drought and predation (mosquitoes and chironomids) and towards tipulids that are relatively resistant to drought, as well as a reduction in secondary production. The effects of drought on the detrital processing chain are harder to predict. Although mosquitoes and chironomids are supressed by direct and indirect effects of drought, they are weak detrital processors compared to tipulids, which are relatively resistant to drought.   This microcosm study has several limitations. Importantly, long-term effects of drought cannot be predicted from our short-term study. Indeed, the intensification of trophic interactions resulting from drought may be short term (Ledger et al. 2013a). For longer droughts, physiological stress on species may outweigh the effect of habitat compression on biotic interactions, and we may expect a decline in mosquitoes and chironomids, as well as tipulids. Furthermore, here we just examine the larval stage of these insects, but changes in climatic variables can also affect adult stages and extrapolation of our results to population-level consequences would require modelling of all the life stages. Apart from studying the long-term effects of drought on bromeliad insect larvae and other life stages, we suggest that future work should focus on manipulating the water column height to study the effects of habitat compression on bromeliad insect communities in more detail.  In conclusion, abiotic change can alter the functional roles of species (e.g., Gormezano and Rockwell 2013; Grémillet et al. 2015), sometimes leading to trophic niche shifts (e.g., McCluney and Sabo 2009; Schmitz et al. 2016; Start et al. 2017). Species that may be particularly likely to mediate indirect effects of climate change by altering their functional role are those with flexible feeding modes or foraging behaviours, such as generalists and omnivores. As such species are ubiquitous across ecosystems (Hawlena et al. 2011; Kratina et al. 2012), this 81  may be a commonly overlooked mechanism by which climate change indirectly affects food webs. Given that a shift in a species’ functional or trophic role will restructure food webs, we need to incorporate potential shifts in how species function under altered environmental conditions in models predicting the ecological effects of climate change.    82  CHAPTER 5   5 DISENTANGLING HOW CLIMATE CHANGE CAN AFFECT AN AQUATIC FOOD WEB BY COMBINING MULTIPLE EXPERIMENTAL APPROACHES4                                                                 4 Previously published as Amundrud, S. L., and Srivastava, D. S. (2019) Disentangling how climate change can affect an aquatic food web by combining multiple experimental approaches. Global Change Biology 25: 3528-3538  83  5.1 CHAPTER SUMMARY  Predicting the biological effects of climate change presents major challenges due to the interplay of potential biotic and abiotic mechanisms. Climate change can create unexpected outcomes by altering species interactions, and uncertainty over the ability of species to develop in situ tolerance or track environmental change further hampers meaningful predictions. As multiple climatic variables shift in concert, their potential interactions further complicate ecosystem responses. Despite awareness of these complexities, we still lack controlled experiments that manipulate multiple climatic stressors, species interactions, and prior exposure of species to future climatic conditions. Particularly studies that address how changes in water availability interact with other climatic stressors to affect aquatic ecosystems are still rare. Using aquatic insect communities of Neotropical tank bromeliads, we combined controlled manipulations of drought length and species interactions with a space-for-time transplant (lower elevations represent future climate) and a common garden approach. Manipulating drought length and experiment elevation revealed that adverse effects of drought were amplified at the warmer location, highlighting the potential of climatic stressors to synergistically affect communities. Manipulating the presence of omnivorous tipulid larvae showed that negative interactions from tipulids, presumably from predation, arose under drought, and were stronger at the warmer location, stressing the importance of species interactions in mediating community responses to climate change. The common garden treatments revealed that prior community exposure to potential future climatic conditions did not affect the outcome. In this powerful experiment, we demonstrated how complexities arise from the interplay of biotic and abiotic mechanisms of climate change. We stress that single species can steer ecological outcomes, and 84  suggest that focusing on such disproportionately influential species may improve attempts at making meaningful predictions of climate change impacts on food webs.   5.2 INTRODUCTION  Climate change will alter communities and ecosystems by affecting both ecological and evolutionary processes (Parmesan 2006), but predicting the biological effects of climate change still presents a major challenge. Most studies have concentrated on the effects of a single climatic variable on one species, ignoring the complexities that arise from the interplay of biotic and abiotic components of climate change. However, climate change affects species not only directly through altering their physiology, behaviour, or life history, but also indirectly by modifying species interactions (Tylianakis et al. 2008; Blois et al. 2013). Such indirect effects of climate change in particular often lead to unexpected or counterintuitive outcomes (Suttle et al. 2007; Ockendon et al. 2014). For instance, altered species interactions resulting from sustained manipulations of precipitation regimes in grassland ecosystems reverse species responses and community trajectories by overriding physiological responses of species to water availability (Suttle et al. 2007). In addition, uncertainty over the ability of species to track environmental change (Tingley et al. 2009) or develop in situ tolerance (Gienapp et al. 2008) makes meaningful predictions ever more difficult. Given the many complexities that arise from climate change, we need controlled experiments that manipulate multiple climatic stressors, species interactions, and prior exposure of species to future climatic conditions. 85  Environmental conditions play major roles in shaping species interactions (Finstad et al. 2011; Hein et al. 2013), and indirect effects of climate change can arise when altered climatic conditions alter species interactions. Asymmetric responses of trophic levels to climate change, for instance, result in trophic mismatches that can disrupt ecosystem functioning by altering top-down or bottom-up processes (Winder and Schindler 2004a; Parain et al. 2016). Alternatively, interaction strengths of predators can be modified by climate change if abiotic conditions alter their functional responses (Hassell 1978; Ewald et al. 2013). Identifying and focusing research efforts on species that are particularly likely to mediate indirect effects of climate change, that is, those that both respond strongly to climate change and have large community level effects, may improve our ability to predict the indirect effects of climate change. Predators in particular have been identified as “biotic multipliers” of climate change (Urban et al. 2017), as they are both responsive to changing climatic conditions and disproportionately affect communities (Amundrud and Srivastava 2016; Urban et al. 2017). Two important aspects of climate change are warming and altered precipitation regimes (IPCC 2014), and both can alter predator-prey interactions. Temperature can affect the functional response of predators, such as by altering their metabolic rates (Rall et al. 2010; Vucic-Pestic et al. 2011), and warming has been shown to alter predation rates in both aquatic (e.g., damselflies: Thompson 1978; fish: Meerhoff et al. 2007) and terrestrial (e.g., spiders: Barton and Schmitz 2009; beetles: Vucic-Pestic et al. 2011) systems. Interestingly, experimental warming has led to both increases and decreases in consumer-resource interactions (Marino et al. 2018). Altered rainfall patterns and increased evaporation rates shorten hydroperiods in aquatic systems (Schindler 2001; Moss 2012), and resulting increases in drought lengths and/or frequencies can affect predator-prey interactions by imposing metabolic and biomechanical constraints on 86  species (Woodward et al. 2016), particularly if different trophic levels are affected asymmetrically (Chase and Knight 2003; Ledger et al. 2013b; Amundrud and Srivastava 2016). As warming and altered precipitation regimes often occur in concert, any interactive effects on communities will further complicate ecosystem responses to climate change. The ecological effects of climate change can further depend on whether species remain vulnerable to altered climatic conditions, or whether they adapt (Skelly et al. 2007), acclimatize (Gienapp et al. 2008), or track (Tingley et al. 2009) the changing climate. Climate change can drive rapid evolution (Harmon et al. 2009a), and phenotypic plasticity can also contribute to phenotypic changes (Gienapp et al. 2008). In addition, species can keep pace with changing climatic conditions by physically tracking their optimal climate (Tingley et al. 2009), such as via range shifts towards higher latitudes (Pinsky et al. 2013) or elevations (Hill et al. 2002). Whether or not species remain naïve to new environmental conditions will likely affect the outcome of climate change, but few studies have incorporated historical exposure to climate into experiments that also manipulate multiple climatic variables and species interactions. What is more, it is improbable that all species in a community respond to climate change in the same way, and asymmetric responses among community members may result in the disruption of food webs (Winder and Schindler 2004a). Any asymmetries in whether or not species remain naïve to new environmental conditions will undoubtedly add further complexities to the effects of climate change. Studying communities along environmental gradients can aid predictions of how they will respond to climate change (Holzapfel et al. 2006; Ariza and Tielbörger 2011), assuming that existing climatic conditions in space mimic predicted climate change in time (Dunne et al. 2004). Altitudinal gradients provide suitable arenas for such space-for-time substitution studies, as 87  considerable climatic gradients occur over short distances. By minimizing the confounding effects of dispersal limitation, which can govern species distributions along latitudinal gradients (Qian and Ricklefs 2007; Amundrud et al. 2018), any turnover in community composition along altitudinal gradients can be attributed to changing abiotic conditions and biotic constraints (Carpenter 2005). However, simply observing community turnover along altitudinal gradients does not permit mechanistic insights, and simultaneous experimental manipulations of abiotic and biotic components are necessary to disentangle the various mechanisms by which climate affects composition and phenotype (Dunne et al. 2004; Liancourt et al. 2013). Community transplant experiments in combination with manipulations of species interactions (Ariza and Tielbörger 2011; Parain et al. 2016) and common garden experiments (De Block et al. 2013; Alexander et al. 2015) are required to unravel the effects of altered environmental conditions, species interactions, and the potential of species to adjust to the altered climate. However, we are not aware of any studies of animal communities that manipulated in a single experiment more than one climatic stressor, species interactions, and prior exposure of species to future climatic conditions. What is more, experimental manipulations of rainfall, an aspect of climate change that is still understudied, in ecosystems whose existence and dynamics depend almost entirely on precipitation, are still exceedingly rare.  Here, in one powerful experiment, we combined controlled manipulations of drought length and species interactions with a space-for-time transplant and a common garden approach (Figure 5.1). As a model system, we chose aquatic insect communities (consisting of the larvae of mosquitoes, chironomids and tipulids) found within Costa Rican phytothelm bromeliads, as it is logistically feasible to replicate, transplant, and experimentally manipulate these simple communities in microcosms (Amundrud and Srivastava 2015, 2016). Bromeliad food webs are 88  strongly shaped by predator-prey interactions (Srivastava 2006) and, like other small water bodies and temporary waters (Collinson et al. 1995; Wellborn et al. 1996), they are particularly vulnerable to climate change (Zotz and Thomas 1999). The high elevation laboratory at the Monteverde Cloud Forest Reserve (MV, elevation: 1520 m) represented the “current climate”, while the low elevation University of Georgia Costa Rica Campus (UGA, elevation: 1100 m) represented the “future climate” for our focal communities (Figure 5.1). The hotter, more variable conditions at UGA reflect future temperature predictions for MV under the A2 (doubled CO2, 2071–2100) scenario (Karmalkar et al. 2008), while the lower humidity at UGA reflects warming in combination with expected declines in mist frequency due to a rise in the cloud base at MV (Pounds et al. 2006; Karmalkar et al. 2008). Our experiment had three different combinations of community origins and experimental locations (expressed in the form Origin→Location, including MV→MV, UGA→UGA, and MV→UGA; Figure 5.1). As such, the space-for-time transplant (MV→UGA) mimicked future climatic conditions, while transplanting high elevation species back into high elevation sites (MV→MV) controlled for current conditions. By accounting for any potential within-species variation arising under different climate histories, the common garden (i.e., communities originating from high [MV→UGA] vs. low [UGA→UGA] elevations) explored whether or not the outcome of simulated climate change depends on prior exposure of species to “future climatic conditions” (i.e., representing situations in which species acclimate to, adapt to, or track climate change). At MV, reductions in precipitation of up to 30 % are predicted under the A2 scenario (Karmalkar et al. 2008), and manipulating drought length allowed us to decouple the effects of resulting increased droughts from the effects of warming. Finally, manipulating the presence of tipulid larvae, which are disproportionately important consumers in terms of litter decomposition 89  (LeCraw 2014) and have recently been discovered to prey on other insects under drought conditions (Amundrud and Srivastava 2016; Amundrud et al. 2019), investigated the role of trophic interactions in mediating any community responses to simulated climate change.    Figure 5. 1: Experimental setup. 24 of the 48 bromeliad-dwelling detritivore communities collected at MV were placed at MV (MV→MV, blue arrow and box), and the other 24 communities were placed at UGA (MV→UGA, green arrow and box). The 24 communities collected at UGA were placed at UGA (UGA→UGA, red arrow and box). For each of these three Origin→Location treatments, we manipulated drought length (12 levels: 0 to 22 days) and tipulid presence (present and absent), resulting in 24 distinct treatment combinations (see magnified blue box for example). Comparing the communities at the two experimental locations thus represents a space-for-time design, while comparing the two communities at UGA with different origins represents a common garden design. Orange box: Temperature at the two locations inside microcosms subjected to drought and water obtained from iButtons. Dark blue: MV-water, red: MV-drought, light blue: UGA-water, orange: UGA-drought.      90  We demonstrate that climatic stressors can interact to affect communities, as the adverse effects of drought on the insect community were amplified at the warmer location. Moreover, we show that indirect effects of climate change mediated by species interactions can outweigh any direct effects, as adverse effect of tipulids on the community arose under drought and were stronger at the warmer location. Finally, we found no evidence that prior exposure of communities to “future climatic conditions” affects the outcome of simulated climate change. Importantly, we demonstrate that a single species can drastically affect how food webs respond to climate change, and support the proposition that predators can be important mediators of climate change (Urban et al. 2017).   5.3 METHODS 5.3.1 Study Sites and System The Monteverde Cloud Forest Reserve entrance (MV: 10°18'N 84°47'W) and the University of Georgia (UGA: 10°16'N 84°47'W) Costa Rica Campus are located on the Pacific slope of the Tilaran mountain range, northwestern Costa Rica. The two sites are ~ 2.5 km apart and differ in altitude by ~ 420 m (Figure 5.1). Tank bromeliads (Vriesea gladioliflora H. Wendl, Guzmania scherzeriana Mez, and Vriesea sanguinolenta Cogn. and Marchal) in this area contain food webs consisting of macroinvertebrates and microorganisms that are largely sustained by detritus from the surrounding trees. The dominant macroinvertebrates in the system are the larvae of dipterans (Tipulidae: undescribed Trentepohlia spp., Chironomidae: especially 91  Polypedilum spp. and Orthocladiinae; and Culicidae: mainly Culex spp., Wyeomyia spp., and Anopheles spp.), as well as coleopteran Scirtidae (Scirtes spp.).   5.3.2 Manipulative Experiment On Oct. 9, 2016, we assembled 48 insect communities collected at MV and 24 identical communities (except for the identity of the Culex spp., see below) collected at UGA in microcosms mimicking single bromeliad leaf wells, the scale at which individual insects spend most time foraging and interacting with each other. We constructed microcosms, which allowed us to precisely manipulate community structure and drought length, using semi-permeable 50 mL centrifuge tubes placed in plastic cups, identical to those used in Amundrud and Srivastava (2015, 2016) and Amundrud et al. (2019). Insect communities were contained in the tubes, which were permeable to water. By adding or removing water in the outer cup, we could manipulate the water level inside the microcosm tubes without disturbing the microcosm contents. To each microcosm, we added 3 mL microbial inoculant in the form of filtered (through 80 µm mesh) water obtained from bromeliads. We added detritus in the form of a pinch of oven-dried detrital particles (80–850 µm diameter) obtained from bromeliads, as well as leaf fragments (~ 1 cm2) of Conostegia xalapensis Bonpl. (Melastomataceae; net dry mass of 54.7 ± 1.2 mg; mean ± SE), which, after drying to obtain initial mass, we rehydrated for three days prior to the experiment. We placed emergence traps on the microcosms, and filled microcosm cups with filtered stream water to a volume of 15 mL inside the tubes (i.e., a water depth of roughly 2.1 cm), which is within the typical range of volume for one bromeliad leaf well (Amundrud and Srivastava 2015). Using filtered stream water ensured that water quality, which varies widely among bromeliads 92  and even leaf wells within the same bromeliad, was standardized among microcosms. Filtered stream water has also commonly been used in other bromeliad food web studies (Srivastava 2006; Starzomski et al. 2010; Amundrud and Srivastava 2015, 2016; Amundrud et al. 2019) and to store live bromeliad insects for many weeks. We added prey communities in the form of 5 mosquito (3 Wyeomyia pertinans and 2 Culex spp.) and 6 chironomid (2 Polypedilum spp. and 4 Orthocladiinae spp.) larvae, and placed tipulid (Trentepohlia spp.) larvae to half of the microcosms (12 at UGA and 24 at MV). The identity of the Culex spp. differed between the two locations (Culex erethyzonfer at MV and Culex jenningsi at UGA), because neither species was abundant enough at both locations, suggesting a degree of prior habitat filtering (Appendix D.1). Although natural bromeliads contain species-rich communities (we recorded an average of 11.9 and 9.5 taxa inside bromeliads at UGA and MV, respectively), the average number of species at the scale of the leaf well is 3-7 (Trzcinski et al. 2016; LeCraw and Srivastava 2019; D. Srivastava, unpubl. data). The number of species per microcosm in our experiment is thus consistent with natural bromeliad systems, and communities in our study represent “typical” bromeliad communities (Appendix D.1). We manipulated climate in a space-for-time design, by placing half of the communities collected at MV (n = 24) at MV (MV→MV), representing the current climate, and the other half (n = 24) at UGA (MV→UGA), representing the future climate (Figure 5.1). At each location, we added one tipulid larva to half of the microcosms (n = 12), and manipulated drought (12 drought lengths ranging from 0 to 22 days, at 2 day intervals; Figure 5.1). Our manipulation of drought was identical to methods employed in previous experiments studying effects of drought on bromeliad microcosm communities (Amundrud and Srivastava 2015, 2016; Amundrud et al. 2019), but to summarize: We initiated drought (i.e., absence of water) in all communities except 93  of those subjected to a drought length of zero days by pouring all water from the microcosm cups (field observations from Costa Rica show that bromeliads lose as much as 1-3 cm of water depth daily; Diane Srivastava, unpubl. data). Although this procedure removed the entire water column from the microcosm tubes, moisture was retained within the leaves and added fine detritus. Manipulating experimental location and drought length allowed us to test for any main and/or interactive effects of simulated future climate and drought on the community, as well as any indirect effects of these aspects of climate change mediated by predation by tipulids. We employed a common garden approach to examine if the outcome of simulated climate change depended on whether or not communities are naïve to altered climatic conditions. To achieve this, we placed the 24 communities collected near UGA at the UGA location (UGA→UGA), subjecting them to the same experimental treatments (Figure 5.1). These communities represented communities already accustomed to a warmer, dryer climate, while the communities originating from MV (MV→UGA) represented communities naïve to future climatic conditions.  Microcosms were placed under a cover outside where temperature variation and evaporation levels were comparable to the adjacent forest, but without water input from precipitation. To monitor temperature within microcosms at the two experimental locations, we prepared six additional microcosms per location (3 with water and 3 without water) that did not include insect communities, but instead held iButtonsTM (DS1921G-F5# Thermochron, 4K) that logged the temperature hourly (Figure 5.1). We obtained relative humidity values from weather stations at the sites from Aug. 25, 2010 to Feb. 7, 2011, the only timeframe that was available for MV. 94  Every two days, we ended the appropriate drought treatments by adding filtered stream water to the respective microcosms to a level of 15 mL inside the tubes. We also replaced 50 % of the water in no-drought microcosms to mimic the natural flushing of bromeliads by rain, and recorded any emergences. On Oct. 31, we emptied microcosms in random order to quantify surviving insects. We oven-dried the detritus and determined decomposition rates over the time of the experiment as the difference in dry mass before and after the experiment.  5.3.3 Statistical Analyses We used logistic regression models with binomial errors to quantify the effects of community type (categorical, 3 levels: MV→MV, MV→UGA, and UGA→UGA), drought length (continuous, 12 levels: 0 to 22 days at 2 day intervals), and tipulid presence (categorical, 2 levels: present and absent), as well as their pairwise interaction (testing for a three-way interaction was not possible due to low statistical power), on prey survival (and chironomid and mosquito survival separately), and linear models to examine treatment effects on prey emergences and decomposition. We calculated prey survival as the fraction of surviving larvae and pupae but excluded the few insects that emerged (Amundrud and Srivastava 2016), as most emergences occurred during the drought treatment and so the intended drought length was not achieved. To tease apart any effects of the Origin→Location treatment on our responses, we employed the same logistic regression models in combination with orthogonal contrasts in cases in which the Origin→Location treatment was significant. We defined orthogonal contrasts as (1) 95  MV→MV vs. MV→UGA and UGA→UGA (space-for-time contrast) and (2) MV→UGA vs. UGA→UGA (common garden contrast; Figure 5.1). We examined model suitability by checking for overdispersion in binomial models, as well as by using quantile-quantile, residual, and outlier plots. We log-transformed drought length for prey survival, and squared drought length for decomposition as the relationship of these responses with drought length were non-linear and transformation improved model fit. Tipulid survival and prey emergences required no transformations of the explanatory variables. We performed all analyses using R statistical software (R Core Team 2017).   5.4 RESULTS  Drought reduced the survival of prey, but the magnitude depended on the Origin→Location treatment (Table 5.1: O→L x D). This interaction was driven more by experimental location than community origin, as when we partitioned the Origin→Location treatment into its two component contrasts, we found that drought interacted with the space-for-time contrast, but not the common garden contrast (Figure 5.2a, Table 5.1: SfT x D). Specifically, the negative effects of drought on total prey survival were stronger at the site representing the future climate (UGA), regardless of community origin (Figure 5.2e). Chironomids drove this response of the total prey community to the drought x space-for-time interaction (Table 5.1: O→L x D and SfT x D, Figure 5.2b). While chironomid survival was unaffected by drought at MV (i.e., the current climate), survival decreased with drought at UGA 96  (Figure 5.2f). In contrast, the negative effects of drought on mosquitoes (Table 5.1: D) were independent of experimental location or whether mosquitoes had been filtered locally by climate (Table 5.1: O→L x D, Figure 5.2c).   97    Table 5. 1: ANOVA results for insect survival (total prey, chironomids, and mosquitoes) and decomposition (as loss of dry mass of detritus). Generalized linear models with Χ2 tests were used for insect survival, and linear models with F tests were used for decomposition. Model predictors are O→L = Origin→Location treatment (MV→MV, MV→UGA, or UGA→UGA), T = Tipulid treatment (present or absent), and D = ln(Drought length + 1). When O→L or its interaction with T or D was significant, orthogonal contrasts decomposed O→L into SfT = Space-for-Time contrast (MV→MV vs. MV→UGA and UGA→UGA) and CG = Common Garden contrast (MV→UGA vs. UGA→UGA).        Predictor   Response   O→L T D O→L x T O→L x D T x D  SfT CG    SfT x T CG x T  SfT x D CG x D  Survival Χ22 P Z P Z P Χ21 P Χ21 P Χ22 P Z P Z P Χ22 P Z P Z P Χ21 P Total prey 3.62 0.164 -- -- -- -- 92.95 <0.001 75.89 <0.001 6.48 0.039 2.51 0.012 0.09 0.932 7.35 0.025 2.69 0.007 0.41 0.679 8.69 0.003 Chironomids 6.81 0.033 -2.18 0.029 -1.03 0.301 37.71 <0.001 14.63 <0.001 3.53 0.171 -- -- -- -- 6.46 0.040 2.48 0.013 0.16 0.875 2.70 0.100 Mosquitoes 0.55 0.762 -- -- -- -- 60.16 <0.001 77.10 <0.001 4.92 0.085 -- -- -- -- 3.70 0.157 -- -- -- -- 14.40 <0.001  F2 P t P t P F1 P F1 P F2 P t P t P F2 P t P t P F1 P Decomposition 4.83 0.011 -1.71 0.092 -0.12 0.907 46.69 <0.001 0.19 0.664 0.35 0.707 -- -- -- -- 0.36 0.698 -- -- -- -- 1.82 0.182    98   Figure 5. 2: Effects of drought length, Origin→Location (O→L) treatment, and tipulid presence on (a) total prey survival, (b) chironimid survival, (c) mosquito survival, and (d) leaf loss. Experimental location represents current (MV) and future (UGA) climate. O→L treatments are indicated by color. Blue: MV→MV, green: MV→UGA, red: UGA→UGA. Tipulid treatment is indicated by solid circles and lines (tipulid present) versus empty circles and dashed lines (tipulid absent). Right box: Partial residuals showing the effect of drought length and experimental location on (e) total prey survival and (f) chironomid survival to illustrate significant drought x space-for-time interaction. Only regression lines of model pooled over the tipulid treatment are shown. Color indicates experimental location. Blue: Monteverde, orange: UGA. Shaded areas represent 95 % confidence intervals.  99    Tipulids also reduced prey survival, and the magnitude depended on the Origin→Location treatment (Table 5.1: O→L x T). This interaction was again driven by experimental location rather than community origin, as tipulids interacted with the space-for-time contrast, but not the common garden contrast (Figure 5.2a, Table 5.1: SfT x T). Specifically, when tipulids were present, prey survival at UGA (i.e., the “future climate”) was 40.6 % lower than at MV (Figure 5.3a), while prey survival did not differ between experimental locations when tipulids were absent (Figure 5.3b).    Figure 5. 3: Partial residuals showing the effect of the space-for-time treatment on total prey survival for communities in which (a) tipulids were present and (b) tipulids were absent. Experimental location is indicated by color. Blue: MV, orange: UGA. Shaded areas represent 95 % confidence intervals.    100   In addition, the adverse effects of tipulids on total prey survival intensified with increasing drought length (Table 5.1: T x D, Figure 5.2a). This tipulid x drought interaction was primarily driven by mosquitoes (Table 5.1, Figure 5.2c). While tipulids had negligible effects when there was no drought (i.e., drought length = 0 days), tipulids significantly decreased total prey and mosquito survival by an average of 83 % and 91 %, respectively, when drought length was two or more days (Figure 5.4). The negative effects of tipulids on chironomid survival, on the other hand, were independent of drought length (Table 5.1, Figure 5.2b).    Figure 5. 4: Partial residuals showing the effect of tipulid treatment on (a-c) total prey survival and (d-f) mosquito survival for each community treatment. Blue: MV→MV, green: MV→UGA, red: UGA→UGA. Tipulid treatment is indicated by solid circles and lines (tipulid present) versus empty circles and dashed lines (tipulid absent). Shaded areas represent 95 % confidence intervals.  101   Decomposition rates increased when tipulids were present, and were affected by the Origin→Location treatment, but not by drought or any treatment interaction (Table 5.1). Decomposition rates were 55.5 % higher when tipulids were present than when they were absent, consistent with tipulids being important shredders of detritus (Figure 5.2d). Moreover, experimental location was more important in affecting decomposition than community origin, as decomposition rates were 24.2 % higher at UGA (i.e., the “future climate”) than at MV (Figure 5.2d). However, this space-for-time effect is only marginally significant (Table 5.1: SfT). Tipulid presence also determined the relationship of prey mortality and litter decomposition rates. While decomposition rates increased with increasing prey mortality in the presence of tipulids, their relationship was negative when tipulids were absent (Appendix D.2).   Prey emergence rates were low (only 8.5 %) and not affected by the space-for-time treatment or the common garden treatment, but the adverse effects of drought on prey emergence rates were stronger when tipulids were present (Appendix D.3). Tipulid survival was not affected by the drought treatment, the space-for-time treatment, the common garden treatment, or any of their potential interactions.    5.5 DISCUSSION  Simulated climate change altered community structure and ecosystem processes in our experiment. The effects on the insect community were primarily mediated by the presence of 102   tipulids, which are known to become predatory under drought (Amundrud et al. 2019). These negative effects of tipulids were accentuated at the warmer location. In addition, direct adverse effects of drought on the community depended on temperature and were stronger at the warmer location. Although decomposition tended to be higher at the warmer location, tipulid presence was more important in affecting decomposition than simulated climate change. Thus, tipulids were disproportionately important in affecting both biotic and abiotic aspects of the bromeliad food web. Through their role as decomposers, tipulids were the main predictor of decomposition rates, and through their potential role as predators, tipulids mediated indirect effects of simulated climate change on the insect community. For chironomids, but not for mosquitoes, the adverse effects of drought depended on temperature, underlining not only that different climatic stressors can interact to affect communities, but also that specific responses to climatic stressors often depend on species identity. Whereas drought reduced chironomid survival only at the warm location, it adversely affected mosquito survival at both locations. These species-specific responses to drought may stem from their different microhabitat use; while mosquitoes depend on vertical movement through the water column to carry out vital functions such as feeding, chironomids can carry out these functions as long as there is moisture. In our experiment, the effect of drought on the state of the water column was independent of location and thus temperature (i.e., the water column was removed at both study sites), explaining why mosquitoes experienced negative effects of drought at both locations. On the other hand, drying rates of the substrate after removal of water were likely location-dependent. The removal of the water column was unlikely to lead to complete drying of the substrate and/or insect bodies at MV, which is submersed in clouds for much of the time. In contrast, low humidity combined with high temperatures at UGA lead to 103   higher evaporation and thus desiccation rates with drought, explaining why chironomids experienced negative effects of drought only at UGA. By disentangling the effects of drought and temperature on the prey community, these results emphasise that different climatic variables often interact to affect organisms on a species-specific basis. As communities consist of diverse species assemblages, the complexities that arise from the multifaceted effects of climate change have to be accounted for when attempting to predict its ecological effects. Altered species interactions further complicated community responses to climate change in our experiment, as both drought and warmer temperatures enhanced the negative effects of tipulids on the prey community. Drought enhanced adverse effects of tipulids on mosquitoes, but not on chironomids, again stressing that climate change responses are often species specific. By removing the water column, drought likely hampered the ability of mosquitoes to escape predation by tipulids, supporting previous conclusions that climate change can inhibit the capability of prey to flee from predators (de Barros et al. 2010; Amundrud et al. 2019). On the other hand, drought is unlikely to significantly alter the mobility of chironomids, which can still avoid predation by moving along the substrate, likely explaining why drought did not enhance negative effects of tipulids on chironomids (Amundrud et al. 2019). Indeed, many chironomids survived in the presence of tipulids even at long drought lengths. Although we were not able to specifically measure prey consumption rates by tipulids in this study due to logistical constraints, we know from previous experiments that drought (through both the removal and the reduction of the water column) enables tipulids to consume insect prey (Amundrud and Srivastava 2016; Amundrud et al. 2019). In one of these previous studies (Amundrud et al. 2019), we used manipulations of caged and non-caged tipulids to differentiate between consumptive and non-consumptive effects of tipulids on other insects, and showed that only consumption was 104   important under drought conditions. Importantly, these consumptive effects of tipulids under drought conditions were particularly severe for mosquitoes, while chironomids were able to avoid predation to some extent, likely by still being mobile in the absence of water. This difference between mosquitoes and chironomids is echoed in our current study. The negative effects of tipulids on the prey community were also enhanced by the warmer temperatures at UGA. Tipulids may have exhibited higher predation rates at the warmer location due to behavioural or metabolic changes caused by higher temperatures (Gilman 2017). Warming has been shown to increase predation rates in a diverse suite of predators such as fish (Meerhoff et al. 2007), terrestrial (Harmon et al. 2009a) and aquatic (De Block et al. 2013) insects. However, a recent meta-analysis found that warming generally weakens top-down control at warm locations such as the tropics (Marino et al. 2018), which is inconsistent with the increase of adverse effects of tipulids with warming in our experiment. Here, warming may have increased predation rates by affecting the predator’s foraging behaviour (Barton 2011) and/or activity levels (Dell et al. 2014). Regardless of the mechanism, our findings suggests that the warmer and dryer climate predicted with climate change will likely affect bromeliad communities through increased predation rates from tipulids, which may translate into increased extinction risk for prey (De Block et al. 2013), and demonstrate that altered abiotic conditions can affect communities by altering the functional role of one of its members. Decomposition rates were primarily affected by tipulid presence, and tended to be higher at the warmer location. Despite the general consensus that climate and litter quality are the best predictors of decomposition rates (Harmon et al. 2009b; Currie et al. 2010), the disproportionate effect of tipulids on litter decomposition in our experiment is consistent with previous findings in 105   this system (Amundrud and Srivastava 2016). Tipulids are known to have the highest per capita litter decomposition rates of all detritivores in tank bromeliads (LeCraw 2014), even after accounting for their relatively large body size (Srivastava and Bell 2009). Interestingly, increased prey consumption by tipulids facilitated by drought and higher temperatures at UGA did not affect their role as decomposers, suggesting that tipulids supplement their diet with prey, rather than switching their diet. A supplementary rather than complementary role of tipulid predation with litter processing is also supported by the positive relationship of decomposition rates with prey mortality in communities where tipulids were present (Appendix D.2), suggesting that those tipulids that exhibit high litter processing rates are also consuming a lot of prey. As decomposition rates were largely independent of temperature and drought, our results support other studies showing that detritivore composition can be more important in influencing litter decomposition than climate (Schimel and Schaeffer 2012; Bradford et al. 2016).  The common garden manipulation represented a strong test of the ability of the insect community to adapt (in the broad sense) to climatic differences. Yet, the origin of communities, whether UGA (i.e., “future” climatic conditions) or MV (i.e., “current” climatic conditions), had no effect on how they responded to simulated climate change. However, tipulids do display a ~ 3 °C higher heat tolerance (i.e., CTmax) at UGA than at MV (Appendix D.4), which corresponds to the actual temperature difference between the locations. Despite this apparent climatic filtering of tipulids, tipulid origin was not important in driving any responses to experimental treatments, nor did tipulids from different origins differentially affect the prey community or decomposition rates. Although the origin of insect larvae in our experiment had no effect on the outcome of simulated climate change, it is possible that filtering by local climate is more important for other life stages (such as adults or eggs). 106   Here, in one powerful experiment, we combined controlled manipulations of drought length and species interactions with a space-for-time transplant and a common garden approach. This design enabled us to decouple the effects of two climatic stressors, warming and drought, as well as account for the role of species interactions in mediating any community responses to these stressors. In addition, the common garden approach allowed us to explore if the outcome of simulated climate change depended on whether communities were locally filtered by climate (representing situations in which communities acclimate to, adapt to, or track climate change). We found that the adverse effects of drought on the bromeliad community were enhanced at the warmer location, adding to the growing amount of evidence that different climatic stressors interactively affect communities (Vinebrooke et al. 2004; Darling et al. 2013). Moreover, adverse effects of simulated climate change on the prey community were amplified by negative effects of tipulids, supporting the view that predators can act as biotic multipliers of climate change (Urban et al. 2017). Importantly, the presence of just one community member, tipulids, drastically affected how the food web responded to climate change, stressing that a single species can steer ecological outcomes to climate change. Focusing predominantly on such “leverage species” (Harley et al. 2006) may improve our ability to predict the ecological outcomes of climate change.    107   CHAPTER 6   6 CONCLUSION    108    This dissertation demonstrates first that spatial scale can influence the relative importance of abiotic and biotic processes in shaping species distributions and ecological communities, and second that the interplay of abiotic and biotic processes can depend on environmental context. I reached these conclusions by combining local experiments with regional surveys and modeling techniques in a natural model system spanning a large geographic range. Through this combination of methods, I was able to disentangle the factors that operate across a range of spatial scales, from the local habitat, to the landscape, to the geographic region. While species interactions were crucial in shaping community composition and ecosystem functions at the local scale, biotic interactions were not important predictors of species distributions at the geographic scale. Interestingly, at the intermediate scale of the landscape, the relative importance of abiotic and biotic processes depended on environmental context, an important insight given that abiotic conditions are already shifting as a result of climate change.  At the geographic scale, physical barriers to dispersal (i.e., ocean barriers and the Andes mountain range) and climatic constraints (i.e., low temperatures and low precipitation) were the main predictors of the distribution of the damselfly M. modesta, while biotic interactions from other bromeliad-dwelling odonates were not important (Chapter 2; Amundrud et al. 2018). The finding that M. modesta is limited by cold and dry conditions is consistent with several other studies. For example, I demonstrated in my Masters’ research that M. modesta exhibits a disproportionately high vulnerability to drought compared to other bromeliad invertebrates, and as a result does not occupy small bromeliads that have a high likelihood of experiencing drought (Amundrud and Srivastava 2015; Srivastava et al. 2020). Second, the observational survey of bromeliad communities at Monteverde Mountain (Chapter 3; Amundrud and Srivastava 2020) 109   demonstrates that this predator is absent from areas that are relatively cold (i.e., high elevations) and dry (i.e., the Pacific slope). Because M. modesta is limited by cold conditions, it is likely that this predator will track climate change to higher elevations, with potentially strong effects on naïve high-elevation bromeliad food webs.   At the landscape scale, the relative importance of abiotic and biotic processes in determining invertebrate elevational distributions and community structure depended on environmental context (Chapter 3; Amundrud and Srivastava 2020). On the wet Atlantic slope, insect physiological tolerances to high temperatures explained their elevational distributions better than trophic interactions from damselflies or crane flies. On the dry Pacific slope, by contrast, negative effects of crane flies on other insects were more important in explaining their elevational distributions than physiological tolerance limits to high temperatures. The finding that damselflies did not exert any significant biotic effects on the community was surprising, given that these predators can exert large-scale community and ecosystem-level effects (e.g., Amundrud and Srivastava 2016), and was likely due to damselflies being rare in areas where they were present (i.e., on the low elevation Atlantic slope). Instead, crane flies, which were disproportionately abundant on the warm and dry low elevation Pacific slope, exerted negative effects on other bromeliad invertebrates and thus mediated the effect of elevation on their distributions. It is likely that access to a more lucrative food source under drought conditions (Chapter 4; Amundrud et al. 2019), together with a relatively high heat tolerance (Chapter 3; Amundrud and Srivastava 2020) and drought tolerance (Amundrud and Srivastava 2015, 2016, 2019) caused crane flies to thrive under dry and warm conditions. These results are also consistent with the findings of an observational study by Melissa Guzman, a former member of 110   the Srivastava lab, who showed that the trophic interaction strength of bromeliad-dwelling crane flies in Brazil increases as bromeliad water level decreases (Guzman et al. 2018). Because crane flies responded strongly to abiotic conditions in terms of their abundance and their functional response (i.e., a drought-induced trophic niche shift), they will likely act as “biotic multiplier” of climate change (Zarnetske et al. 2012; Urban et al. 2017) when conditions become warmer and dryer (Karmalkar et al. 2008). These findings shed new light on complex community-level responses to climate change, which are still poorly understood. Not only do they support the idea that the effects of climate change on species ranges and community dynamics at the landscape scale may be highly influenced by species interactions (Clark et al. 2011; Diamond et al. 2016), but they also demonstrate that the interplay of abiotic and biotic selective pressures on species physiology and distributions can be highly dynamic and depend on environmental context. This is an important insight given the rapid rates of changing environmental conditions as a result of global change.  At the local scale, the effects of simulated climate change on the invertebrate community were strongly mediated by biotic interactions from crane flies, whose presence altered both community structure and ecosystem processes far beyond any direct community-level effects of warming or drought. At this scale, negative interactions from crane flies on other insects arose under drought conditions (Chapters 4 and 5; Amundrud and Srivastava 2019; Amundrud et al. 2019), and were accentuated at the warmer location (Chapter 5; Amundrud and Srivastava 2019). As such, the presence of just one community member drastically affected how the food web responded to simulated climate change. Moreover, this work uncovered the mechanism of how simulated climate change resulted in crane flies exerting such strong effects on the community. 111   Using controlled experiments, I demonstrated that crane flies became predatory under drought conditions, and that this drought-induced predatory behavior arose because of increased encounter rates of crane flies with potential prey species as a result of habitat compression. As such, climate change can alter the trophic structure of a food web affecting the functional role of an opportunistic consumer. Because opportunistic foragers are relatively common across ecosystems (Hawlena et al. 2011; Kratina et al. 2012), this may be a commonly overlooked mechanism by which climate change indirectly affects food webs.  I took advantage of a broad-ranging natural ecosystem that can be easily manipulated and replicated to learn about the processes that govern communities and species distributions across a range of spatial scales. By combining modeling techniques with observational surveys along environmental gradients and controlled experiments, I demonstrated that the relative importance of abiotic and biotic processes that affect species distributions and the structure of ecological communities can be scale-dependent. While biotic interactions were key drivers of community responses at small and intermediate spatial scales, they were not detectable at the large geographical scale, lending support to the Eltonian Noise Hypothesis (Soberón and Nakamura 2009). Interestingly, at the intermediate scale of the landscape, the importance of biotic interactions in driving species elevational distributions and thus community structure depended on environmental context, suggesting that, at the landscape scale, the relative importance of physiological and biological drivers of species distributions may shift with changing abiotic conditions such as climate change. These results were driven by the response of a single species to abiotic conditions, and caused by the trophic niche shift of the crane fly larvae in response to drought. Altered top-down effects from climate change will likely affect food webs across many 112   ecosystem types (Marino et al. 2018), but studies on how potential shift in species’ functional or trophic roles under altered environmental conditions will restructure food webs are still scarce. Fresh waters are particularly vulnerable to climate change, and in several areas of the world the prevalence or duration of drought events is likely to increase (Prudhomme et al. 2014). It is thus possible that habitat compression in ponds or streams due to drought will lead to similar effects on increased top down control in food webs (Dudgeon 1993). Future studies that investigate the dynamic nature of the relative importance of abiotic and biotic effects in other systems would be informative.   As ecologists hasten to predict how communities and ecosystems will respond to global change, many efforts still focus on direct responses of species to altered environmental conditions, ignoring complexities that arise from altered species interactions. However, the evidence that species interactions can be of paramount importance in mediating the effects of global change on ecological communities (Suttle et al. 2007) and ecosystem processes (Schmitz et al. 2003) is growing. However, due to the sheer number of complex interactions among species and their environment, such indirect effects are inherently difficult to predict, and little progress has been made in incorporating species interactions into ecological forecast models. Harley et al. (2006) proposed that impacts on “leverage species” can result in large-scale ecosystem level changes (Harley et al. 2006). However, the challenge remains to correctly identify such leverage species. Based on the findings of my PhD and MSc theses, potential leverage species might be species that exhibit a strong response to environmental change (e.g., through demographic and/or per capita responses), as well as having disproportionately large ecosystem-level impacts (e.g., through exerting strong interactions on other species). For 113   example, in my MSc work I showed that disproportionately adverse effects of drought on the survival of a keystone predator damselfly larva can reshape communities and alter ecosystem functions far beyond any changes resulting from direct effects of drought on individual species (Amundrud and Srivastava 2016). In this dissertation, I found that crane flies responded to drought by shifting their trophic role, causing them to exert strong consumptive effects on other insects with subsequent community-level effects that exceeded any direct effects of drought. Future research could synthesize the literature to explore whether species that mediate effects of global change display certain traits that make them likely to respond strongly to environmental change and/or have large community level effects. By focusing on characteristics that determine why certain species might be particularly likely to mediate global change effects, such a mechanistic framework could be used to help (1) identify (and direct conservation efforts towards) species that are particularly likely to mediate the effects of global change, (2) predict which communities are most likely to experience indirect effects, and (3) improve predictions of how ecosystems respond to global change.  Climate change is rapidly altering food webs globally, but we still lack an adequate understanding of how global change affects food webs and their associated ecosystem functions. Apart from the challenge to incorporate species interactions into forecast models, there are additional aspects of multifaceted global change that hamper our ability to make accurate predictions, such as multiple climatic variables shifting in concert with other drivers of global change such as land use, and uncertainty over the ability of species to develop in situ tolerance or track environmental change. I demonstrated how we can unravel some of these complexities, by performing the first controlled experiment that manipulated multiple climatic stressors, species 114   interactions, and prior exposure of species to future climatic conditions, and thus showed how uncertainties arise from the interplay of multiple biotic and abiotic mechanisms of climate change (Amundrud and Srivastava 2019). However, we still only have a limited understanding of how climate change and other aspects of global change interact to affect food webs and their associated ecosystem functions. Future studies that combine observational surveys along land use and climate gradients with controlled experiments that manipulate climate and land use (e.g., space-for-time transplants, warming experiments in the field, etc.) and species interactions (presence or absence of potential leverage species) could be a good first step in uncovering the combined effects of climate change and land use change on ecological systems.      115    BIBLIOGRAPHY  Addo-Bediako A, Chown SL, Gaston KJ (2000) Thermal tolerance, climatic variability and latitude. Proc R Soc B 267:739–745 . doi: 10.1098/rspb.2000.1065 Alexander JM, Diez JM, Levine JM (2015) Novel competitors shape species’ responses to climate change. Nature 525:515–518 . doi: 10.1038/nature14952 Amundrud SL, Clay-Smith SA, Flynn BL, et al (2019) Drought alters the trophic role of an opportunistic generalist in an aquatic ecosystem. Oecologia 189:733–744 . doi: 10.1007/s00442-019-04343-x Amundrud SL, Srivastava DS (2016) Trophic interactions determine the effects of drought on an aquatic ecosystem. Ecology 97:1475–1483 . doi: 10.1890/15-1638.1 Amundrud SL, Srivastava DS (2015) Drought sensitivity predicts habitat size sensitivity in an aquatic ecosystem. Ecology 96:1957–1965 . doi: 10.1890/14-1828.1 Amundrud SL, Srivastava DS (2019) Disentangling how climate change can affect an aquatic food web by combining multiple experimental approaches. Glob Chang Biol. doi: 10.1111/gcb.14717 Amundrud SL, Srivastava DS (2020) Thermal tolerances and species interactions determine the elevational distributions of insects. Glob Ecol Biogeogr 29:1315–1327 . doi: 10.1111/geb.13106 Amundrud SL, Videla M, Srivastava DS (2018) Dispersal barriers and climate determine the geographic distribution of the helicopter damselfly Mecistogaster modesta. Freshw Biol 63:214–223 . doi: 10.1111/fwb.13054 Ancona S, Calixto-Albarrán I, Drummond H (2012) Effect of El Niño on the diet of a specialist seabird, Sula nebouxii, in the warm eastern tropical Pacific. Mar Ecol Prog Ser 462:261–271 . doi: 10.3354/meps09851 Antonelli A, Nylander JAA, Persson C, et al (2009) Testing the impact of the Andean uplift on Neotropical plant evolution. Proc Natl Acad Sci 106:9749–9754 . doi: 10.1073/pnas.081 Araújo MB, Luoto M (2007) The importance of biotic interactions for modelling species distributions under climate change. Glob Ecol Biogeogr 16:743–753 . doi: 10.1111/j.1466-116   8238.2007.00359.x Ariza C, Tielbörger K (2011) An evolutionary approach to studying the relative importance of plant-plant interactions along environmental gradients. Funct Ecol 25:932–942 . doi: 10.1111/j.1365-2435.2011.01848.x Atwood TB, Hammill E, Greig HS, et al (2013) Predator-induced reduction of freshwater carbon dioxide emissions. Nat Geosci 6:191–194 . doi: 10.1038/ngeo1734 Atwood TB, Hammill E, Srivastava DS, Richardson JS (2014) Competitive displacement alters top-down effects on carbon dioxide concentrations in a freshwater ecosystem. Oecologia 175:353–61 . doi: 10.1007/s00442-013-2877-3 Báez S, Collins SL, Lightfoot D, Koontz TL (2006) Bottom-up regulation of plant community structure in an aridland ecosystem. Ecology 87:2746–54 Baker RL, Ball SL (1995) Microhabitat selection by larval Chironomus tentans (Diptera: Chironomidae): effects of predators, food, cover and light. Freshw Biol 34:101–106 . doi: 10.1111/j.1365-2427.1995.tb00427.x Ball-Damerow JE, M’Gonigle LK, Resh VH (2014a) Changes in occurrence, richness, and biological traits of dragonflies and damselflies (Odonata) in California and Nevada over the past century. Biodivers Conserv 23:2107–2126 . doi: 10.1007/s10531-014-0707-5 Ball-Damerow JE, M’Gonigle LK, Resh VH (2014b) Local and regional factors influencing assemblages of dragonflies and damselflies (Odonata) in California and Nevada. J Insect Conserv 18:1027–1036 . doi: 10.1007/s10841-014-9709-6 Barton BT (2011) Local adaptation to temperature conserves top-down control in a grassland food web. Proc Biolocial Sci 278:3102–3107 . doi: 10.1098/rspb.2011.0030] Barton BT, Schmitz OJ (2009) Experimental warming transforms multiple predator effects in a grassland food web. Ecol Lett 12:1317–25 . doi: 10.1111/j.1461-0248.2009.01386.x Baudier KM, D’Amelio CL, Malhotra R, et al (2018) Extreme insolation: Climatic variation shapes the evolution of thermal tolerance at multiple scales. Am Nat 192:347–359 . doi: 10.1086/698656 Benke AC (1978) Interactions among coexisting predators - a field experiment with dragonfly larvae. J Anim Ecol 47:335–350 . doi: 10.2307/3787 Benzing DH (2000) Bromeliaceae: Profile of an adaptive radiation. Cambridge University Press, Cambridge Blois JL, Zarnetske PL, Fitzpatrick MC, Finnegan S (2013) Climate change and the past, present, and future of biotic interactions. Science (80- ) 341:499–504 . doi: 10.1126/science.1237184 Boukal DS (2014) Trait- and size-based descriptions of trophic links in freshwater food webs: 117   current status and perspectives. J Limnol 73:171–185 . doi: 10.4081/jlimnol.2014.826 Boulangeat I, Gravel D, Thuiller W (2012) Accounting for dispersal and biotic interactions to disentangle the drivers of species distributions and their abundances. Ecol Lett 15:584–593 . doi: 10.1111/j.1461-0248.2012.01772.x Boyd PW, Cornwall CE, Davison A, et al (2016) Biological responses to environmental heterogeneity under future ocean conditions. Glob Chang Biol 22:2633–2650 . doi: 10.1111/gcb.13287 Bozinovic F, Calosi P, Spicer JI (2011) Physiological correlates of geographic range in animals. Annu Rev Ecol Evol Syst 42:155–179 . doi: 10.1146/annurev-ecolsys-102710-145055 Brackenbury J (2000) Locomotory modes in the larva and pupa of Chironomus plumosus (Diptera, Chironomidae). J Insect Physiol 46:1517–1527 . doi: 10.1016/S0022-1910(00)00079-2 Brackenbury J (2001) Locomotion through use of the mouth brushes in the larva of Culex pipiens (Diptera: Culicidae). Proc R Soc B Biol Sci 268:101–106 . doi: 10.1098/rspb.2000.1336 Bradford MA, Berg B, Maynard DS, et al (2016) Understanding the dominant controls on litter decomposition. J Ecol 104:229–238 . doi: 10.1111/1365-2745.12507 Brattstrom B (1968) Thermal acclimation in anuran amphibians as a function of latitude and altitude. Comp Biochem Physiol 24:93–111 . doi: 10.1016/0010-406X(68)90961-4 Braune E, Richter O, Söndgerath D, Suhling F (2008) Voltinism flexibility of a riverine dragonfly along thermal gradients. Glob Chang Biol 14:470–482 . doi: 10.1111/j.1365-2486.2007.01525.x Brodie J, Post E, Laurance WF (2012) Climate change and tropical biodiversity: A new focus. Trends Ecol Evol 27:145–150 . doi: 10.1016/j.tree.2011.09.008 Bruno JF, Stachowicz JJ, Bertness MD (2003) Inclusion of facilitation into ecological theory. Trends Ecol Evol 18:119–125 . doi: 10.1016/S0169-5347(02)00045-9 Buchanan JA, Stewart BA, Davies BR (1988) Thermal acclimation and tolerance to lethal high temperature in the mountain stream amphipod Paramelita nigroculus (Barnard). Comp Biochem Physiol 89A:425–431 . doi: 10.1016/0300-9629(88)91051-1 Cargill A, Cummins K, Hanson B, Lowry R (1985) The role of lipids as feeding stimulants for shreeding aquatic insects. Freshw Biol 15:455–464 . doi: 10.1111/j.1365-2427.1985.tb00215.x Carnicer J, Abrams PA, Jordano P (2008) Switching behavior, coexistence and diversification: Comparing empirical community-wide evidence with theoretical predictions. Ecol Lett 11:802–808 . doi: 10.1111/j.1461-0248.2008.01195.x 118   Carpenter C (2005) The environmental control of plant species density on a Himalayan elevation gradient. J Biogeogr 32:999–1018 . doi: 10.1111/j.1365-2699.2005.01249.x Carroll AL, Taylor SW, Régnière J, Safranyik L (2003) Effects of climate change on range expansion by the mountain pine beetle in British Columbia. In: Mountain pine beetle symposium: Challenges and solutions. Kelowna, British Columbia, pp 223–232 Chase JM, Knight TM (2013) Scale-dependent effect sizes of ecological drivers on biodiversity: why standardised sampling is not enough. Ecol Lett 16:17–26 . doi: 10.1111/ele.12112 Chase JM, Knight TM (2003) Drought-induced mosquito outbreaks in wetlands. Ecol Lett 6:1017–1024 . doi: 10.1046/j.1461-0248.2003.00533.x Christophers R (1960) Aëdes aegypti (L.), the yellow fever mosquito: Its life history, bionomics and structure. Cambridge University Press, New York Clark JS, Bell DM, Hersh MH, Nichols L (2011) Climate change vulnerability of forest biodiversity: Climate and competition tracking of demographic rates. Glob Chang Biol 17:1834–1849 . doi: 10.1111/j.1365-2486.2010.02380.x Clusella-Trullas S, Blackburn TM, Chown SL (2011) Climatic predictors of temperature performance curve parameters in ectotherms imply complex responses to climate change. Am Nat 177:738–751 . doi: 10.1086/660021 Coll M, Hughes L (2008) Effects of elevated CO2 on an insect omnivore: a test for nutritional effects mediated by host plants and prey. Agric Ecosyst Environ 123:271–279 . doi: 10.1016/j.agee.2007.06.003 Collins SD, Heintzman LJ, Starr SM, et al (2014) Hydrological dynamics of temporary wetlands in the southern Great Plains as a function of surrounding land use. J Arid Environ 109:6–14 . doi: 10.1016/j.jaridenv.2014.05.006 Collins SD, McIntyre NE (2015) Modeling the distribution of odonates: a review. Freshw Sci 34:1144–1158 . doi: 10.1086/682688 Collinson NH, Biggs J, Corfield A, et al (1995) Temporary and permanent ponds: An assessment of the effects of drying out on the conservation value of aquatic macroinvertebrate communities. Biol Conserv 74:125–133 . doi: 10.1016/0006-3207(95)00021-U Connell JH (1980) Diversity and the coevolution of competitors, or the ghost of competition past. Oikos 35:131–138 . doi: 10.2307/3544421 Cooper BS, Williams BH, Angilletta MJ (2008) Unifying indices of heat tolerance in ectotherms. J Therm Biol 33:320–323 . doi: 10.1016/j.jtherbio.2008.04.001 Costa HC, de Rezende DT, Molina FB, et al (2015) New distribution records and potentially suitable areas for the threatened snake-necked turtle Hydromedusa maximiliani (Testudines: Chelidae). Chelonian Conserv Biol 14:88–94 . doi: 10.2744/ccab-14-01-88-94.1 119   Currie WS, Harmon ME, Burke IC, et al (2010) Cross-biome transplants of plant litter show decomposition models extend to a broader climatic range but lose predictability at the decadal time scale. Glob Chang Biol 16:1744–1761 . doi: 10.1111/j.1365-2486.2009.02086.x Darling ES, McClanahan TR, Côté IM (2013) Life histories predict coral community disassembly under multiple stressors. Glob Chang Biol 19:1930–1940 . doi: 10.1111/gcb.12191 Davis A, Jenkinson L, Lawton J, et al (1998) Making mistakes when predicting shifts in species range in response to global warming. Nature 391:783–786 . doi: 10.1038/35842 De-Silva DL, Elias M, Willmott K, et al (2016) Diversification of clearwing butterflies with the rise of the Andes. J Biogeogr 43:44–58 . doi: 10.1111/jbi.12611 De Araújo CB, Marcondes-Machado LO, Costa GC (2014) The importance of biotic interactions in species distribution models: A test of the Eltonian noise hypothesis using parrots. J Biogeogr 41:513–523 . doi: 10.1111/jbi.12234 de Barros FC, de Carvalho JE, Abe AS, Kohlsdorf T (2010) Fight versus flight: the interaction of temperature and body size determines antipredator behaviour in tegu lizards. Anim Behav 79:83–88 . doi: 10.1016/j.anbehav.2009.10.006 De Block M, Pauwels K, Van Den Broeck M, et al (2013) Local genetic adaptation generates latitude-specific effects of warming on predator-prey interactions. Glob Chang Biol 19:689–696 . doi: 10.1111/gcb.12089 De Marco P, Latini AO, Resende DC (2005) Thermoregulatory constraints on behavior: patterns in a Neotropical dragonfly assemblage. Neotrop Entomol 34:155–162 . doi: 10.1590/S1519-566X2005000200002 Dell AI, Pawar S, Savage VM (2014) Temperature dependence of trophic interactions are driven by asymmetry of species responses and foraging strategy. J Anim Ecol 83:70–84 . doi: 10.1111/1365-2656.12081 Deutsch CA, Tewksbury JJ, Huey RB, et al (2008) Impacts of climate warming on terrestrial ectotherms across latitude. Proc Natl Acad Sci 105:6668–6672 . doi: 10.1073/pnas.0709472105 Diamond SE, Nichols LM, Pelini SL, et al (2016) Climatic warming destabilizes forest ant communities. Sci Adv 2: . doi: 10.1126/sciadv.1600842 Dillon ME, Frazier MR, Dudley R (2006) Into thin air: Physiology and evolution of alpine insects. Integr Comp Biol 46:49–61 . doi: 10.1093/icb/icj007 Dingemanse NJ, Kalkman VJ (2008) Changing temperature regimes have advanced the phenology of Odonata in the Netherlands. Ecol Entomol 33:394–402 . doi: 10.1111/j.1365-2311.2007.00982.x 120   Domisch S, Jähnig SC, Haase P (2011) Climate-change winners and losers: Stream macroinvertebrates of a submontane region in Central Europe. Freshw Biol 56:2009–2020 . doi: 10.1111/j.1365-2427.2011.02631.x Duarte H, Tejedo M, Katzenberger M, et al (2012) Can amphibians take the heat? Vulnerability to climate warming in subtropical and temperate larval amphibian communities. Glob Chang Biol 18:412–421 . doi: 10.1111/j.1365-2486.2011.02518.x Dudgeon D (1993) The effects of spate-induced disturbance, predation and environmental complexity on macroinvertebrates in a tropical stream. Freshw Biol 30:189–197 . doi: 10.1111/j.1365-2427.1993.tb00801.x Dunne JA, Saleska SR, Fischer ML, Harte J (2004) Integrating experimental and gradient methods in ecological climate change research. Ecology 85:904–916 . doi: 10.1890/03-8003 Emmons LH, Feer F (1997) Neotropical rainforest mammals: A field guide, Second Edi. The University of Chicago Press, Chicago Endler JA (1982) Problems in distinguishing historical from ecological factors in biogeography. Am Zool 22:441–452 . doi: 10.1093/icb/22.2.441 Ernest SK, Brown JH, Parmenter RR (2000) Rodents, plants, and precipitation: spatial and temporal dynamics of consumers and resources. Oikos 88:470–482 Ewald NC, Hartley SE, Stewart AJA (2013) Climate change and trophic interactions in model temporary pond systems: The effects of high temperature on predation rate depend on prey size and density. Freshw Biol 58:2481–2493 . doi: 10.1111/fwb.12224 Fincke OM (1994) Population regulation of a tropical damselfly in the larval stage by food limitation, cannibalism, intraguild predation and habitat drying. Oecologia 100:118–127 . doi: 10.1007/BF00317138 Finstad AG, Forseth T, Jonsson B, et al (2011) Competitive exclusion along climate gradients: Energy efficiency influences the distribution of two salmonid fishes. Glob Chang Biol 17:1703–1711 . doi: 10.1111/j.1365-2486.2010.02335.x Freeman EA, Moisen G (2008) PresenceAbsence: An R package for presence absence analysis. J Stat Softw 23:1–31 García-Robledo C, Kuprewicz EK, Staines CL, et al (2016) Limited tolerance by insects to high temperatures across tropical elevational gradients and the implications of global warming for extinction. PNAS 113:680–685 . doi: 10.1073/pnas.1507681113 Gaston KJ, Chown SL (1999) Elevation and climatic tolerance: A test using dung beetles. Oikos 86:584–590 . doi: 10.2307/3546663 Gestoso I, Arenas F, Olabarria C (2015) Feeding behaviour of an intertidal snail: does past environmental stress affect predator choices and prey vulnerability? J Sea Res 97:66–74 . doi: 10.1016/j.seares.2014.12.006 121   Giannini TC, Chapman DS, Saraiva AM, et al (2013) Improving species distribution models using biotic interactions: A case study of parasites, pollinators and plants. Ecography (Cop) 36:649–656 . doi: 10.1111/j.1600-0587.2012.07191.x Gienapp P, Teplitsky C, Alho JS, et al (2008) Climate change and evolution: Disentangling environmental and genetic responses. Mol Ecol 17:167–178 . doi: 10.1111/j.1365-294X.2007.03413.x Gilman SE (2017) Predicting indirect effects of predator-prey interactions. Integr Comp Biol 57:148–158 . doi: 10.1093/icb/icx031 Givnish TJ, Spalink D, Ames M, et al (2016) Orchid historical biogeography, diversification, Antarctica and the paradox of orchid dispersal. J Biogeogr 43:1905–1916 . doi: 10.1111/jbi.12854 Godsoe W, Franklin J, Blanchet FG (2017a) Effects of biotic interactions on modeled species’ distribution can be masked by environmental gradients. Ecol Evol 7:654–664 . doi: 10.1002/ece3.2657 Godsoe W, Jankowski J, Holt RD, Gravel D (2017b) Integrating biogeography with contemporary niche theory. Trends Ecol Evol 32:488–499 . doi: 10.1016/j.tree.2017.03.008 Gormezano LJ, Rockwell RF (2013) What to eat now? Shifts in polar bear diet during the ice-free season in western Hudson Bay. Ecol Evol 3:3509–3523 . doi: 10.1002/ece3.740 Graham CH, Carnaval AC, Cadena CD, et al (2014) The origin and maintenance of montane diversity: Integrating evolutionary and ecological processes. Ecography (Cop) 37:711–719 . doi: 10.1111/ecog.00578 Graham CH, Parra JL, Rahbek C, McGuire JA (2009) Phylogenetic structure in tropical hummingbird communities. Proc Natl Acad Sci 106:19673–19678 . doi: 10.1073/pnas.0901649106 Grémillet D, Fort J, Amélineau F, et al (2015) Arctic warming: nonlinear impacts of sea-ice and glacier melt on seabird foraging. Glob Chang Biol 21:1116–1123 . doi: 10.1111/gcb.12811 Guzman LM, Vanschoenwinkel B, Farjalla VF, et al (2018) A precipitation gradient drives change in macroinvertebrate composition and interactions within bromeliads. PLoS One 13:1–14 . doi: 10.1371/journal.pone.0200179 Hammill E, Atwood TB, Corvalan P, Srivastava DS (2015a) Behavioural responses to predation may explain shifts in community structure. Freshw Biol 60:125–135 . doi: 10.1111/fwb.12475 Hammill E, Atwood TB, Srivastava DS (2015b) Predation threat alters composition and functioning of bromeliad ecosystems. Ecosystems 18:857–866 Harley CDG, Hughes AR, Hultgren KM, et al (2006) The impacts of climate change in coastal marine systems. Ecol Lett 9:228–241 . doi: 10.1111/j.1461-0248.2005.00871.x 122   Harmon JP, Moran NA, Ives AR (2009a) Species response to environmental change: Impacts of food web interactions and evolution. Science (80- ) 323:1347–1351 . doi: 10.1126/science.1167396 Harmon ME, Silver WL, Fasth B, et al (2009b) Long-term patterns of mass loss during the decomposition of leaf and fine root litter: An intersite comparison. Glob Chang Biol 15:1320–1338 . doi: 10.1111/j.1365-2486.2008.01837.x Hassall C, Thompson DJ, French GC, Harvey IF (2007) Historical changes in the phenology of British Odonata are related to climate. Glob Chang Biol 13:933–941 . doi: 10.1111/j.1365-2486.2007.01318.x Hassell MP (1978) The dynamics of arthropod predator-prey systems. Princeton University Press, Princeton, NJ Hawlena D, Hughes KM, Schmitz OJ (2011) Trophic trait plasticity in response to changes in resource availability and predation risk. Funct Ecol 25:1223–1231 . doi: 10.1111/j.1365-2435.2011.01891.x Hedström I, Sahlén G (2001) A key to the adult Costa Rican “helicopter” damselflies (Odonata: Pseudostigmatidae) with notes on their phenology and life zone preferences. Rev Biol Trop 49:1037–1056 Heikkinen RK, Luoto M, Virkkala R, et al (2007) Biotic interactions improve prediction of boreal bird distributions at macro-scales. Glob Ecol Biogeogr 16:754–763 . doi: 10.1111/j.1466-8238.2007.00345.x Hein CL, Ohlund G, Englund G (2013) Fish introductions reveal the temperature dependence of species interactions. Proc R Soc B Biol Sci 281:20132641–20132641 . doi: 10.1098/rspb.2013.2641 Hernandez PA, Graham C, Master LL, Albert DL (2006) The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography (Cop) 29:773–785 . doi: 10.1111/j.0906-7590.2006.04700.x Hijmans RJ, Cameron SE, Parra JL, et al (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Climatol 25:1965–1978 . doi: 10.1002/joc.1276 Hill JK, Thomas CD, Fox R, et al (2002) Responses of butterflies to twentieth century climate warming: implications for future ranges. Proc R Soc B Biol Sci 269:2163–2171 . doi: 10.1098/rspb.2002.2134 Hochachka P, Somero G (2002) Biochemical adaptation: mechanism and process in physiological evolution. Oxford University Press, New York Hoffmann A, Anderson A, Hallas R (2002) Opposing clines for high and low temperature resistance in Drosophila melanogaster. Ecol Lett 5:614–618 . doi: 10.1046/j.1461-0248.2002.00367.x 123   Holt RD, Polis GA (1997) A theoretical framework for intraguild predation. Am Nat 149:745–764 . doi: 10.1086/286018 Holzapfel C, Tielbörger K, Parag HA, et al (2006) Annual plant-shrub interactions along an aridity gradient. Basic Appl Ecol 7:268–279 . doi: 10.1016/j.baae.2005.08.003 Humphries MM, Studd EK, Menzies AK, Boutin S (2017) To everything there is a season: Summer-to-winter food webs and the functional traits of keystone species. Integr Comp Biol 57:961–976 . doi: 10.1093/icb/icx119 Ingley SJ, Bybee SM, Tennessen KJ, et al (2012) Life on the fly: Phylogenetics and evolution of the helicopter damselflies (Odonata, Pseudostigmatidae). Zool Scr 41:637–650 . doi: 10.1111/j.1463-6409.2012.00555.x IPCC (2014) Climate Change 2014: Synthesis Report Jacobsen D (2003) Altitudinal changes in diversity of macroinvertebrates from small streams in the Ecuadorian Andes. Arch für Hydrobiol 158:145–167 . doi: 10.1127/0003-9136/2003/0158-0145 Jha S (2015) Contemporary human-altered landscapes and oceanic barriers reduce bumble bee gene flow. Mol Ecol 24:993–1006 . doi: 10.1111/mec.13090 Johnson DM, Crowley PH, Bohanan RE, et al (1985) Competition among larval dragonflies : A field enclosure experiment. Ecology 66:119–128 . doi: 10.2307/1941312 Jørgensen LB, Malte H, Overgaard J (2019) How to assess Drosophila heat tolerance: Unifying static and dynamic tolerance assays to predict heat distribution limits. Funct Ecol 33:629–642 . doi: 10.1111/1365-2435.13279 Karmalkar A V., Bradley RS, Diaz HF (2008) Climate change scenario for Costa Rican montane forests. Geophys Res Lett 35:L11702 . doi: 10.1029/2008GL033940 Kaspari M, Clay NA, Lucas J, et al (2015) Thermal adaptation generates a diversity of thermal limits in a rainforest ant community. Glob Chang Biol 21:1092–1102 . doi: 10.1111/gcb.12750 Kearney M, Porter W (2009) Mechanistic niche modelling: combining physiological and spatial data to predict species’ ranges. Ecol Lett 12:334–50 . doi: 10.1111/j.1461-0248.2008.01277.x Kellermann V, Overgaard J, Hoffmann AA, et al (2012) Upper thermal limits of Drosophila are linked to species distributions and strongly constrained phylogenetically. Proc Natl Acad Sci 109:16228–16233 . doi: 10.1073/pnas.1207553109 Khazan ES (2014) Tests of biological corridor efficacy for conservation of a Neotropical giant damselfly. Biol Conserv 177:117–125 . doi: 10.1016/j.biocon.2014.06.006 Kivivuori LA, Lahdes EO (1996) How to measure the thermal death of Daphnia? A comparison 124   of different heat tests and effects of heat injury. J Therm Biol 21:305–311 . doi: 10.1016/S0306-4565(96)00014-9 Kratina P, LeCraw RM, Ingram T, Anholt BR (2012) Stability and persistence of food webs with omnivory: is there a general pattern? Ecosphere 3:50 . doi: 10.1890/ES12-00121.1 Kruse PD, Toft S, Sunderland KD (2008) Temperature and prey capture: opposite relationships in two predator taxa. Ecol Entomol 33:305–312 . doi: 10.1111/j.1365-2311.2007.00978.x Lawton JH, Hassell MP (1981) Asymmetrical competition in insects. Nature 289:793–795 . doi: 10.1038/289793a0 LeCraw RM (2014) The influences of spatial processes at multiple scales on local community structure and function. (Doctoral Dissertation). University of British Columbia. Retrieved from http://hdl.handle.net/2429/45901 LeCraw RM, Srivastava DS (2019) Biogeographic context dependence of trophic cascade strength in bromeliad food webs. Ecology in press:e02692 . doi: 10.1002/ecy.2692 Ledger ME, Brown LE, Edwards FK, et al (2013a) Extreme climatic events alter aquatic food webs. A synthesis of evidence from a mesocosm drought experiment. Adv Ecol Res 48:343–395 . doi: 10.1016/B978-0-12-417199-2.00006-9 Ledger ME, Brown LE, Edwards FK, et al (2013b) Drought alters the structure and functioning of complex food webs. Nat Clim Chang 3:223–227 . doi: 10.1038/nclimate1684 Ledger ME, Edwards FK, Brown LE, et al (2011) Impact of simulated drought on ecosystem biomass production: an experimental test in stream mesocosms. Glob Chang Biol 17:2288–2297 . doi: 10.1111/j.1365-2486.2011.02420.x Ledger ME, Harris RML, Armitage PD, Milner AM (2012) Climate change impacts on community resilience: evidence from a drought disturbance experiment. Adv Ecol Res 46:211–258 . doi: 10.1016/B978-0-12-396992-7.00003-4 Ledger ME, Harris RML, Armitage PD, Milner AM (2008) Disturbance frequency influences patch dynamics in stream benthic algal communities. Oecologia 155:809–19 . doi: 10.1007/s00442-007-0950-5 Lemoine NP (2017) Predation risk reverses the potential effects of warming on plant-herbivore interactions by altering the relative strengths of trait- and density-mediated interactions. Am Nat 190:337–349 . doi: 10.1086/692605 Liancourt P, Spence LA, Song DS, et al (2013) Plant response to climate change varies with topography, interactions with neighbors, and ecotype. Ecology 94:444–453 . doi: 10.1890/12-0780.1 Lutterschmidt WI, Hutchison VH (1997) The critical thermal maximum: history and critique. Can J Zool 75:1561–1574 125   Lutz PE (1968) Effects of temperature and photoperiod on larval development in Lestes eurinus (Odonata: Lestidae). Ecology 49:637–644 . doi: 10.2307/1935224 MacAvoy SE, Braciszewski A, Tengi E, Fong DW (2016) Trophic plasticity among spring vs. cave populations of Gammarus minus: examining functional niches using stable isotopes and C/N ratios. Ecol Res 31:589–595 . doi: 10.1007/s11284-016-1359-6 Macchiusi F, Baker RL (1992) Effects of predators and food availability on activity and growth of Chironomus tentans (Chironomidae, Diptera). Freshw Biol 28:207–216 MacDonald GM (2003) Biogeography: Space, time and life, 6th ed. Wiley, New York Magoulick DD, Kobza RM (2003) The role of refugia for fishes during drought : A review and synthesis. Fres 48:1186–1198 . doi: 10.1046/j.1365-2427.2003.01089.x Maindonald JH, Braun WJ (2015) DAAG: Data analysis and graphics data and functions Marino N, Romero GQ, Farjalla VF (2018) Geographical and experimental contexts modulate the effect of warming on top-down control: a meta-analysis. Ecol Lett 21:455–466 . doi: 10.1111/ele.12913 Marino NAC, Srivastava DS, MacDonald AAM, et al (2017) Rainfall and hydrological stability alter the impact of top predators on food web structure and function. Glob Chang Biol 23:673–685 . doi: 10.1111/gcb.13399 Martin MM, Martin JS, Kukor JJ, Merritt RW (1980) The digestion of protein and carbohydrate by the stream detritivore, Tipula abdominalis (Diptera, Tipulidae). Oecologia 46:360–364 . doi: 10.1007/BF00346265 McCauley SJ, Davis CJ, Werner EE, Robeson MS (2014) Dispersal, niche breadth and population extinction: Colonization ratios predict range size in North American dragonflies. J Anim Ecol 83:858–865 . doi: 10.1111/1365-2656.12181 McCluney KE (2017) Implications of animal water balance for terrestrial food webs. Curr Opin Insect Sci 23:13–21 . doi: 10.1016/j.cois.2017.06.007 McCluney KE, Sabo JL (2009) Water availability directly determines per capita consumption at two trophic levels. Ecology 90:1463–9 . doi: 10.1890/08-1626.1 McGill BJ, Mittelbach GC (2006) An allometric vision and motion model to predict prey encounter rates. Evol Ecol Res 8:691–701 McGuffin MA, Baker RL, Forbes MR (2006) Detection and avoidance of fish predators by adult Enallagma damselflies. J Insect Behav 19:77–91 . doi: 10.1007/s10905-005-9013-0 McHugh PA, Thompson RM, Greig HS, et al (2015) Habitat size influences food web structure in drying streams. Ecography (Cop) 38:700–712 . doi: 10.1111/ecog.01193 McKechnie AE, Wolf BO (2010) Climate change increases the likelihood of catastrophic avian 126   mortality events during extreme heat waves. Biol Lett 6:253–6 . doi: 10.1098/rsbl.2009.0702 McMeans BC, McCann KS, Humphries M, et al (2015) Food web structure in temporally-forced ecosystems. Trends Ecol Evol 30:662–672 . doi: 10.1016/j.tree.2015.09.001 Meerhoff M, Clemente JM, de Mello FT, et al (2007) Can warm climate-related structure of littoral predator assemblies weaken the clear water state in shallow lakes? Glob Chang Biol 13:1888–1897 . doi: 10.1111/j.1365-2486.2007.01408.x Merritt RW, Cummins KW (1995) Habitat, life history, and behavioral adaptations of aquatic insects. In: Merritt RW, Cummins KW (eds) An introduction to the aquatic insects of North America, 3rd edn. Kendall Hunt Pub Co, Dubuque, Iowa, pp 41–73 Montoya JM, Woodward G, Emmerson MC, Solé R V (2009) Press perturbations and indirect effects in real food webs. Ecology 90:2426–33 Moss B (2012) Cogs in the endless machine: Lakes, climate change and nutrient cycles: A review. Sci Total Environ 434:130–142 . doi: 10.1016/j.scitotenv.2011.07.069 Ngai JT, Srivastava DS (2006) Predators accelerate nutrient cycling in a bromeliad ecosystem. Science (80- ) 314:963 . doi: 10.1126/science.1132598 Nicholson AJ (1958) Dynamics of insect populations. Annu Rev Entomol 3:107–136 . doi: 10.1146/annurev.en.03.010158.000543 Nowrouzi S, Andersen AN, Bishop TR, Robson SKA (2018) Is thermal limitation the primary driver of elevational distributions? Not for montane rainforest ants in the Australian Wet Tropics. Oecologia 188:333–342 . doi: 10.1007/s00442-018-4154-y O’Connor MI (2009) Warming strengthens an herbivore-plant interaction. Ecology 90:388–398 . doi: 10.1890/08-0034.1 Ockendon N, Baker DJ, Carr JA, et al (2014) Mechanisms underpinning climatic impacts on natural populations: altered species interactions are more important than direct effects. Glob Chang Biol 20:2221–9 . doi: 10.1111/gcb.12559 Overgaard J, Kearney MR, Hoffmann AA (2014) Sensitivity to thermal extremes in Australian Drosophila implies similar impacts of climate change on the distribution of widespread and tropical species. Glob Chang Biol 20:1738–1750 . doi: 10.1111/gcb.12521 Oyen KJ, Giri S, Dillon ME (2016) Altitudinal variation in bumble bee (Bombus) critical thermal limits. J Therm Biol 59:52–57 . doi: 10.1016/j.jtherbio.2016.04.015 Parain EC, Gravel D, Rohr RP, et al (2016) Mismatch in microbial food webs: predators but not prey perform better in their local biotic and abiotic conditions. Ecol Evol 6:4885–4897 . doi: 10.1002/ece3.2236 Parmesan C (2006) Ecological and evolutionary responses to recent climate change. Annu Rev 127   Ecol Evol Syst 37:637–669 . doi: 10.1146/annurev.ecolsys.37.091305.110100 Pearson RG, Dawson TP (2003) Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Glob Ecol Biogeogr 12:361–371 . doi: 10.1046/j.1466-822X.2003.00042.x Pearson RG, Raxworthy CJ, Nakamura M, Peterson AT (2007) Predicting species distributions from small numbers of occurrence records: A test case using cryptic geckos in Madagascar. J Biogeogr 34:102–117 . doi: 10.1111/j.1365-2699.2006.01594.x Petchey OL, Brose U, Rall BC (2010) Predicting the effects of temperature on food web connectance. Philos Trans R Soc B Biol Sci 365:2081–2091 . doi: 10.1098/rstb.2010.0011 Petermann JS, Farjalla VF, Jocque M, et al (2015) Dominant predators mediate the impact of habitat size on trophic structure in bromeliad invertebrate communities. Ecology 96:428–439 . doi: 10.1890/14-0304.1 Peterson AT, Soberón J, Pearson RG, et al (2011) Ecological niches and geographic distributions. Princeton University Press, Princeton, N.J. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecol Modell 190:231–259 . doi: 10.1016/j.ecolmodel.2005.03.026 Pickup J, Thompson DJ (1990) The effects of temperature and prey density on the development rates and growth of damselfIy larvae (Odonata: Zygoptera). Ecol Entomol 15:187–200 . doi: 10.1111/j.1365-2311.1990.tb00800.x Pimm SL, Lawton JH (1978) On feeding on more than one trophic level. Nature 275:542–544 Pinsky ML, Worm B, Fogarty MJ, et al (2013) Marine taxa track local climate velocities. Science (80- ) 341:1239–1242 . doi: 10.1126/science.1239352 Pires APF, Marino NAC, Srivastava DS, Farjalla VF (2016) Predicted rainfall changes disrupt trophic interactions in a tropical aquatic ecosystem. Ecology 97:2750–2759 . doi: 10.1002/ecy.1501 Polis GA, Myers CA, Holt RD (1989) The ecology and evolution of intraguild predation: Potential competitors that eat each other. Annu Rev Ecol Syst 20:297–330 . doi: 10.1146/annurev.es.20.110189.001501 Pounds JA, Bustamante MR, Coloma L a, et al (2006) Widespread amphibian extinctions from epidemic disease driven by global warming. Nature 439:161–167 . doi: 10.1038/nature04246 Power ME (1992) Top-down and bottom-up forces in food webs: do plants have primacy. Ecology 73:733–746 Preisser EL, Bolnick DI, Benard MF (2005) Scared to death? The effects of intimidation and consumption in predator-pery interactions. Ecology 86:501–509 . doi: 10.1890/04-0719 128   Pritchard G (1985) On the locomotion of cranefly larvae (Tipulidae; Tipulinae). J Kansas Entomol Soc 58:152–156 Prudhomme C, Giuntoli I, Robinson EL, et al (2014) Hydrological droughts in the 21st century, hotspots and uncertainties from a global multimodel ensemble experiment. Proc Natl Acad Sci U S A 111:3262–7 . doi: 10.1073/pnas.1222473110 Qian H, Ricklefs RE (2007) A latitudinal gradient in large-scale beta diversity for vascular plants in North America. Ecol Lett 10:737–744 . doi: 10.1111/j.1461-0248.2007.01066.x R Core Team (2015) R: A Language and Environment for Statistical Computing R Core Team (2019) R: A Language and Environment for Statistical Computing R Core Team (2017) R: A Language and Environment for Statistical Computing Rall BC, Vucic-Pestic O, Ehens RB, et al (2010) Temperature, predator-prey interaction strength and population stability. Glob Chang Biol 16:2145–2157 . doi: 10.1111/j.1365-2486.2009.02124.x Richter O, Suhling F, Müller O, Kern D (2008) A model for predicting the emergence of dragonflies in a changing climate. Freshw Biol 53:1868–1880 . doi: 10.1111/j.1365-2427.2008.02012.x Rosseel Y (2012) lavaan: An R package for structural equation modeling. J Sta 48: Samejima Y, Tsubaki Y (2010) Body temperature and body size affect flight performance in a damselfly. Behav Ecol Sociobiol 64:685–692 . doi: 10.1007/s00265-009-0886-3 Sánchez-Herrera M, Ware JL (2012) Biogeography of dragonflies and damselflies: highly mobile predators. In: Stevens L (ed) Global Advances in Biogeography. InTech, pp 291–306 Sanford E (1999) Regulation of keystone predation by small changes in ocean temperature. Science (80- ) 283:2095–2097 . doi: 10.1126/science.283.5410.2095 Sato M, Kawaguchi Y, Yamanaka H, et al (2010) Predicting the spatial distribution of the invasive piscivorous chub (Opsariichthys uncirostris uncirostris) in the irrigation ditches of Kyushu, Japan: A tool for the risk management of biological invasions. Biol Invasions 12:3677–3686 . doi: 10.1007/s10530-010-9762-3 Saupe EE, Barve V, Myers CE, et al (2012) Variation in niche and distribution model performance: The need for a priori assessment of key causal factors. Ecol Modell 237–238:11–22 . doi: 10.1016/j.ecolmodel.2012.04.001 Schimel JP, Schaeffer SM (2012) Microbial control over carbon cycling in soil. Front Microbiol 3:1–11 . doi: 10.3389/fmicb.2012.00348 Schindler DW (2001) The cumulative effects of climate warming and other human stresses on 129   Canadian freshwaters in the new millennium. Can J Fish Aquat Sci 58:18–29 . doi: 10.1139/f00-179 Schmitz OJ, Post E, Burns CE, Johnston KM (2003) Ecosystem responses to global climate change: moving beyond color mapping. Bioscience 53:1199–1205 . doi: 10.1641/0006-3568(2003)053[1199:ERTGCC]2.0.CO;2 Schmitz OJ, Rosenblatt AE, Smylie M (2016) Temperature dependence of predation stress and the nutritional ecology of a generalist herbivore. Ecology 97:3119–3130 . doi: 10.1002/ecy.1524 Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to ImageJ: 25 years of image analysis. Nat Methods 9:671–675 . doi: 10.1038/nmeth.2089 Schoenfeld D (1980) Chi-squared goodness-of-fit tests for the proportional hazards regression model. Biometrika 67:145–153 . doi: Chi-squared goodness-of-fit tests for the proportional hazards regression model Scrucca L (2018) dispmod: Modelling Dispersion in GLM Sentis A, Hemptinne JL, Brodeur J (2014) Towards a mechanistic understanding of temperature and enrichment effects on species interaction strength, omnivory and food-web structure. Ecol Lett 17:785–793 . doi: 10.1111/ele.12281 Silvertown J (2004) The ghost of competition past in the phylogeny of island endemic plants. J Ecol 92:168–173 . doi: 10.1111/j.1365-2745.2004.00853.x Skelly DK, Joseph LN, Possingham HP, et al (2007) Evolutionary responses to climate change. Conserv Biol 21:1353–1355 . doi: 10.1111/j.1523-1739.2007.00764.x Smith JMB (2009) Dispersal of plants and animals to oceanic islands. EOLSS Publications, Oxford Soberón J (2007) Grinnellian and Eltonian niches and geographic distributions of species. Ecol Lett 10:1115–1123 . doi: 10.1111/j.1461-0248.2007.01107.x Soberón J, Nakamura M (2009) Niches and distributional areas: Concepts, methods, and assumptions. Proc Natl Acad Sci 106:19644–19650 . doi: 10.1073/pnas.0901637106 Srivastava DS (2006) Habitat structure, trophic structure and ecosystem function: interactive effects in a bromeliad-insect community. Oecologia 149:493–504 . doi: 10.1007/s00442-006-0467-3 Srivastava DS, Bell T (2009) Reducing horizontal and vertical diversity in a foodweb triggers extinctions and impacts functions. Ecol Lett 12:1016–28 . doi: 10.1111/j.1461-0248.2009.01357.x Srivastava DS, Ware JL, Ngai JT, et al (2020) Habitat size thresholds for predators : Why damselflies only occur in large bromeliads. Biotropica 1–11 . doi: 10.1111/btp.12734 130   Start D, Kirk D, Shea D, Gilbert B (2017) Cannibalism by damselflies increases with rising temperature. Biol Lett 13:20170175 . doi: 10.1098/rsbl.2017.0175 Starzomski BM, Suen D, Srivastava DS (2010) Predation and facilitation determine chironomid emergence in a bromeliad-insect food web. Ecol Entomol 35:53–60 . doi: 10.1111/j.1365-2311.2009.01155.x Sunday JM, Bates AE, Dulvy NK (2011) Global analysis of thermal tolerance and latitude in ectotherms. Proc R Soc B Biol Sci 278:1823–1830 . doi: 10.1098/rspb.2010.1295 Sunday JM, Bates AE, Dulvy NK (2012) Thermal tolerance and the global redistribution of animals. Nat Clim Chang 2:686–690 . doi: 10.1038/nclimate1539 Suttle AKB, Thomsen MA, Power ME (2007) Species interactions reverse grassland responses to changing climate. Science (80- ) 315:640–642 . doi: 10.1126/science.1136401 Suutari E, Rantala MJ, Salmela J, Suhonen J (2004) Intraguild predation and interference competition on the endangered dragonfly Aeshna viridis. Oecologia 140:135–139 . doi: 10.1007/s00442-004-1559-6 Terblanche JS, Deere JA, Clusella-Trullas S, et al (2007) Critical thermal limits depend on methodological context. Proc R Soc B Biol Sci 274:2935–2942 . doi: 10.1098/rspb.2007.0985 Terry M. Therneau (2015) coxme: Mixed Effects Cox Models Therneau T (2018) deming: Deming, Theil-Sen, Passing-Bablock and Total Least Squares Regression Thompson DJ (1978) Towards a realistic predator-prey model: the effect of temperature on the functional response and life history of larvae of the damselfly, Ischnura elegans. J Anim Ecol 47:757–767 . doi: 10.2307/3669 Tingley MW, Monahan WB, Beissinger SR, Moritz C (2009) Birds track their Grinnellian niche through a century of climate change. Proc Natl Acad Sci 106:19637–19643 . doi: 10.1073/pnas.0901562106 Törnroos A, Nordström MC, Aarnio K, Bonsdorff E (2015) Environmental context and trophic trait plasticity in a key species, the tellinid clam Macoma balthica L. J Exp Mar Bio Ecol 472:32–40 . doi: 10.1016/j.jembe.2015.06.015 Traill LW, Lim MLM, Sodhi NS, Bradshaw CJA (2010) Mechanisms driving change: Altered species interactions and ecosystem function through global warming. J Anim Ecol 79:937–947 . doi: 10.1111/j.1365-2656.2010.01695.x Trzcinski MK, Srivastava DS, Corbara B, et al (2016) The effects of food web structure on ecosystem function exceeds those of precipitation. J Anim Ecol 85:1147–1160 . doi: 10.1111/1365-2656.12538 131   Tunney TD, McCann KS, Lester NP, Shuter BJ (2012) Food web expansion and contraction in response to changing environmental conditions. Nat Commun 3:1105–1109 . doi: 10.1038/ncomms2098 Tylianakis JM, Didham RK, Bascompte J, Wardle DA (2008) Global change and species interactions in terrestrial ecosystems. Ecol Lett 11:1351–1363 . doi: 10.1111/j.1461-0248.2008.01250.x Urban MC, Zarnetske PL, Skelly DK (2017) Searching for biotic multipliers of climate change. Integr Comp Biol 57:134–147 . doi: 10.1093/icb/icx045 van Uitregt VO, Hurst TP, Wilson RS (2013) Greater costs of inducible behavioural defences at cooler temperatures in larvae of the mosquito, Aedes notoscriptus. Evol Ecol 27:13–26 . doi: 10.1007/s10682-012-9576-0 Vasseur D, DeLong J, Gilbert B, et al (2014) Increased temperature variation poses a greater risk to species than climate warming. Proc R Soc B 281:1–8 Vázquez DP, Gianoli E, Morris WF, Bozinovic F (2017) Ecological and evolutionary impacts of changing climatic variability. Biol Rev 92:22–42 . doi: 10.1111/brv.12216 Vinebrooke RD, Cottingham KL, Norberg J, et al (2004) Impacts of multiple stressors on biodiversity and ecosystem functioning: The role of species co-tolerance. Oikos 104:451–457 . doi: 10.1111/j.0030-1299.2004.13255.x von Oheimb PV, Albrecht C, Riedel F, et al (2013) Testing the role of the Himalaya Mountains as a dispersal barrier in freshwater gastropods (Gyraulus spp.). Biol J Linn Soc 109:526–534 . doi: 10.1111/bij.12068 Vorhees AS, Gray EM, Bradley TJ (2012) Thermal resistance and performance correlate with climate in populations of a widespread mosquito. Physiol Biochem Zool 86:73–81 . doi: 10.1086/668851 Vucic-Pestic O, Ehnes RB, Rall BC, Brose U (2011) Warming up the system: higher predator feeding rates but lower energetic efficiencies. Glob Chang Biol 17:1301–1310 . doi: 10.1111/j.1365-2486.2010.02329.x Wellborn GA, Skelly DK, Werner EE (1996) Mechanisms creating community structure across a freshwater habitat gradient. Annu Rev Ecol Syst 27:337–363 . doi: 10.1146/annurev.ecolsys.27.1.337 Wellenreuther M, Larson KW, Svensson EI (2012) Climatic niche divergence or conservatism? Environmental niches and range limits in ecologically similar damselflies. Ecology 93:1353–1366 . doi: 10.1890/11-1181.1 Wieczorek J, Guo Q, Hijmans R (2004) The point-radius method for georeferencing locality descriptions and calculating associated uncertainty. Int J Geogr Inf Sci 18:745–767 . doi: 10.1080/13658810412331280211 132   Winder M, Schindler DE (2004a) Climate change uncouples trophic interactions in an aquatic ecosystem. Ecology 85:2100–2106 . doi: 10.1890/04-0151 Winder M, Schindler DE (2004b) Climatic effects on the phenology of lake processes. Glob Chang Biol 10:1844–1856 . doi: 10.1111/j.1365-2486.2004.00849.x Wisz MS, Hijmans RJ, Li J, et al (2008) Effects of sample size on the performance of species distribution models. Divers Distrib 14:763–773 . doi: 10.1111/j.1472-4642.2008.00482.x Wolkovich EM, Cook BI, Allen JM, et al (2012) Warming experiments underpredict plant phenological responses to climate change. Nature 485:494–497 . doi: 10.1038/nature11014 Woodward G, Bonada N, Brown LE, et al (2016) The effects of climatic fluctuations and extreme events on running water ecosystems. Philos Trans R Soc B Biol Sci 371: . doi: 10.1098/rstb.2015.0274 Woodward G, Hildrew AG (2001) Invasion of a stream food web by a new top predator. J Anim Ecol 70:273–288 . doi: 10.1111/j.1365-2656.2001.00497.x Worthen WB, Horacek HJ (2015) The distribution of dragonfly larvae in a South Carolina stream: Relationships with sediment type, body size, and the presence of other larvae. J Insect Sci 15:1–7 . doi: 10.1093/jisesa/iev013 Yanoviak SP (1999) Community ecology of water-filled tree holes in Panama. PhD Diss. Graduate College, University of Oklahoma. Zarnetske PL, Skelly DK, Urban MC (2012) Biotic multipliers of climate change. Science (80- ) 336:1516–8 . doi: 10.1126/science.1222732 Zotz G, Thomas V (1999) How much water is in the tank? Model calculations for two epiphytic bromeliads. Ann Bot 83:183–192 . doi: 10.1006/anbo.1998.0809    133    APPENDICES     134   APPENDIX A   A SUPPLEMENTARY INFORMATION TO CHAPTER 25                                                                   5 Previously published as Supporting Information of Amundrud S.A., Videla M., & Srivastava D.S. (2017) Dispersal barriers and climate determine the geographic distribution of the helicopter damselfly Mecistogaster modesta. Freshwater Biology 63: 214-223.  135   A.1 DETAILED METHODS AND RESULTS OF THE MAXENT MODELS OF THREE POTENTIAL COMPETITORS OF M. MODESTA   To examine whether competition determines the geographic distribution of Mecistogaster modesta, we used the modeled distributions of three of its potential competitors. We searched for occurrence records of these species using the same resources as for M. modesta (see main article). We focused on bromeliad-dwelling odonate species with known occurrence records in the north of South America, as their ranges may overlap with that of M. modesta. The possible competitor species of M. modesta belong to the Bromeliagrion genus and have been recorded in Guyana, Venezuela, Ecuador, Peru, and Brazil. As occurrence records for Bromeliagrion bebeanum, B. fernandezianum and B. rehni were sparse (6, 6, and 11, respectively), we used Maxent, a widely used algorithm for presence only data that performs better than others when sample sizes are small (Hernandez et al. 2006; Wisz et al. 2008), to model their distributions. For each potential competitor species, we ran a Maxent model using 19 bioclimatic variables at 2.5 arc-minutes resolution from the WorldClim database (Hijmans et al. 2005). Based on variable importance and low correlation (r < 0.8), we chose three to four bioclimatic variables to model the distributions of the three odonate species (Table A.1).   We validated models using the jackknife (or “leave one out”) approach proposed by Pearson et al.(2007). This approach is especially suitable when records are sparse and partitioning the data randomly into ‘training’ and ‘test’ sets is not recommended (Pearson et al. 2007). This procedure requires the transformation of continuous scores (suitability index) into binary (presence/absence) data, which is achieved by setting a threshold above which the model output is considered a prediction of presence. Selection of a threshold is controversial given that 136   the choice of threshold could influence the results. Therefore, we performed the analysis using three different threshold selection criteria (minimum training presence, 10% training presence, and equal sensitivity and specificity) following De Araújo et al. (2014). The results did not vary with the threshold selection criteria used, so we present the results obtained from using the equal sensitivity and specificity criteria (Table A.1 and Figure A.1).        Table A. 1: Environmental variables and their importance used in Maxent models of Bromeliagrion species distributions.  Success rates and p-values result from model validation using the leave one out approach (Pearson et al. 2007). Species  BioClim Variable  Contribution  (%) Success  Rate p-value  Bromeliagrion bebeanum   Temperature Seasonality (bio4) 74.6 0.833 0.000061 Precipitation  Seasonality (bio15) 23.0   Precipitation of Coldest Quarter (bio19)  2.4   Bromeliagrion fernandezianum Temperature Seasonality (bio4) 51.9 0.666 0.00185 Temperature Annual Range (bio7) 28.2   Precipitation of Wettest Month (bio13) 16.5   Precipitation Seasonality (bio15)  3.3   Bromeliagrion rehni Temperature Seasonality (bio4)  37.6 0.909 0.000001 Minimal Temperature of Coldest Month (bio6) 35.8   Precipitation of Driest Month (bio14) 24.8   Precipitation Seasonality (bio15) 1.7     137     Figure A. 1: Habitat suitability maps for (a) Bromeliagrion bebeanum, (b) B. fernandezianum, (c) B. rehni and (d) all species combined.         138   A.2 REFERENCES  De Araújo, C. B., Marcondes-Machado, L. O., & Costa, G. C. (2014). The importance of biotic interactions in species distribution models: A test of the Eltonian noise hypothesis using parrots. Journal of Biogeography, 41(3), 513–523. https://doi.org/10.1111/jbi.12234 Hernandez, P. A., Graham, C., Master, L. L., & Albert, D. L. (2006). The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography, 29(5), 773–785. https://doi.org/10.1111/j.0906-7590.2006.04700.x Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25(15), 1965–1978. https://doi.org/10.1002/joc.1276 Pearson, R. G., Raxworthy, C. J., Nakamura, M., & Peterson, A. T. (2007). Predicting species distributions from small numbers of occurrence records: A test case using cryptic geckos in Madagascar. Journal of Biogeography, 34(1), 102–117. https://doi.org/10.1111/j.1365-2699.2006.01594.x Wisz, M. S., Hijmans, R. J., Li, J., Peterson, A. T., Graham, C. H., Guisan, A., … Zimmermann, N. E. (2008). Effects of sample size on the performance of species distribution models. Diversity and Distributions, 14(5), 763–773. https://doi.org/10.1111/j.1472-4642.2008.00482.x        139   APPENDIX B   B SUPPLEMENTARY INFORMATION TO CHAPTER 36                                                               6 Previously published as Supporting Information of Amundrud, S. L., and Srivastava, D. S. (2020) Thermal tolerances and species interactions determine the elevational distributions of insects. Global Ecology and Biogeography 29: 1315-1327  140   B.1  INFORMATION OF THE STUDY SITES ON THE PACIFIC AND ATLANTIC SLOPES OF MONTEVERDE MOUNTAIN   We surveyed bromeliad food webs at nine sites along two elevational gradients (i.e., on the Pacific and Atlantic slope of Monteverde Mountain) from late September to early December in 2015 and 2016, that is, during the rainy season when bromeliads contain most insect larvae (Table B.1). Because sampling time can affect abundance patterns of insects due to seasonal effects, we ensured that sampling occurred in the shortest timeframe that was logistically possible, and was performed non-systematically with respect to mountainside and elevation (Figure B.1). Only two sites (San Luis and Research Trail) were relatively easily accessible, and were thus sampled several times throughout the seasons. The other sites were either only accessible by day hikes (Brillantes and Dos Ases) or multiday field trips (El Valle, San Gerardo, Aleman, Eladios, Poco Sol), and could not be sampled repeatedly in the same year because of logistical constraints.       141     Table B. 1: Mountainside, elevation (in meters), geographic coordinates, and sampling dates of the nine study sites on Monteverde Mountain. The number of bromeliads sampled at each date is indicated in brackets.  site side elevation coordinates sampling dates San Luis Pacific 1082 N10° 16.9' W84° 47.6' 11/10/2015 (n=4); 16/10/2015 (n=4); 13/11/2015 (n=4); 02/12/2015 (n=4);  02/10/2016 (n=1); 03/10/2016 (n=2); 05/10/2016 (n=1); 15/11/2016 (n=1) Research Trail Pacific 1517 N10° 17.8' W84° 47.6' 09/10/2015 (n=3); 15/10/2015 (n=4); 10/11/2015 (n=4); 04/12/2015 (n=3);  04/10/2016 (n=3); 09/10/2016 (n=2)  Brillantes Pacific 1573 N10° 17.5' W84° 47.0' 27/11/2015 (n=5); 31/11/2015 (n=5);  27/11/2016 (n=4) El Valle Atlantic 1590 N10° 19.4' W84° 45.9' 29/09/2015 (n=2); 30/09/2015 (n=6); 01/10/2015 (n=2);  13/10/2016 (n=2); 14/10/2016 (n=3) Dos Ases Atlantic 1379 N10° 17.9' W84° 46.8' 13/10/2015 (n=6); 23/11/2015 (n=6); 25/11/2015 (n=6) San Gerardo Atlantic 1221 N10° 21.8' W84° 47.4' 16/11/2015 (n=4); 17/11/2015 (n=4); 18/11/2015 (n=8); 19/11/2015 (n=3);  22/10/2016 (n=2); 23/10/2016 (n=1)  Aleman Atlantic 974 N10° 18.0' W84° 44.9' 06/10/2015 (n=4); 07/10/2015 (n=7); 08/10/2015 (n=2);  17/10/2016 (n=2); 18/10/2016 (n=3) Eladios Atlantic 809 N10° 18.5' W84° 43.0' 27/10/2015 (n=1); 28/10/2015 (n=5); 29/10/2015 (n=3);  28/10/2016 (n=4); 29/10/2016 (n=1)  Poco Sol Atlantic 787 N10° 21.0' W84° 40.1' 19/10/2015 (n=4); 20/10/2015 (n=5); 21/10/2015 (n=5); 22/10/2015 (n=6); 23/10/2015 (n=1);  03/11/2016 (n=2); 04/11/2016 (n=2)      142      Figure B. 1: Sampling dates of bromeliad communities along elevational gradients on the Pacific (purple crosses) and Atlantic (green circles) slopes of Monteverde Mountain in 2015 and 2016. In both years, sampling was conducted non-systematically with respect to elevation and mountainside.     143   B.2  ELEVATION PREFERENCE AND MEAN ELEVATION OF OCCURRENCE FOR BROMELIAD INSECT SPECIES FOR EACH MOUNTAINSIDE  Because elevation differentially affected insect abundances, elevation preferences (i.e., the slope of the elevation term in the Poisson model; Figure 3.2 in Chapter 3) of bromeliad macroinvertebrates were species-specific. On both mountainsides, marsh beetles (Scirtid spp.) exhibited a preference for high elevations, while chironomids (Polypedilum spp.) showed no significant elevation preference. Damselflies (M. modesta) preferred low elevations at the Atlantic slope, and were absent from the Pacific slope. Interestingly, the elevation preferences of mosquitoes (Culex erethyzonfer) and crane flies (Trentepohlia spp.) depended on mountainside: while mosquitoes preferred high elevations on both slopes, the preference was much stronger on the Atlantic slope (Figure B.2a) than on the Pacific slope (Figure B.2b; elevation preference is marginally non-significant). Crane flies exhibited no elevation preference on the Atlantic slope (Figure B.2a), but they preferred low elevations on the Pacific slope (Figure B.2b).   The mean elevation of occurrence (MEO) of all individuals within each species follows a similar pattern to the elevation preferences of bromeliad insects (Figure B.3). On the Atlantic slope, mosquitoes (C. erethyzonfer) occurred at the highest elevations, damselflies (M. modesta) occurred at the lowest elevations, and the other species occurred at intermediate elevations (Figure B.3a). On the Pacific slope, crane flies (Trentepohlia spp.) occurred at the lowest elevations, while the other three species occurred at similar elevations (Figure B.3b).     144       Figure B. 2: Elevation preferences of bromeliad insects on (a) the Atlantic slope and (b) the Pacific slope of Monteverde Mountain. Elevation preference (EP) is the slope of the elevation term from the Poisson model in Figure 3.2 in Chapter 3. Error bars represent 95% confidence intervals (approximated by 1.96 * SE). When mountainside did not interact with elevation (Polypedilum spp. and Scirtid spp.), the EPs from the model without side is shown. Otherwise, the EPs from separate models for the Atlantic slope and Pacific slope are shown.      145       Figure B. 3: Mean elevation of occurrence (MEO) of each species for insects from (a) the Atlantic slope and (b) the Pacific slope of Monteverde Mountain. Error bars represent standard deviation. Species with large error bars thus occurred widely across elevations, while species with narrow error bars occurred over narrow elevational distributions within our study area.    146   B.3  STRUCTURAL EQUATION MODELS USING AN OBLIGATE PREDATOR VARIABLE INSTEAD OF THE M. MODESTA VARIABLE AS POTENTIAL MEDIATOR TO EXPLAIN EFFECTS OF ELEVATION ON FOCAL SPECIES   When replacing M. modesta with an obligate predator variable [i.e., by using the combined abundances of diving beetles (Copelatus sp.) and damselflies (M. modesta)], the results are qualitatively similar to using only damselflies as in Chapter 3 (Table B.2).    Table B. 2: The relative abundance of each of three focal species (columns) is modelled with structural equation models (SEMs), with each SEM being composed of multiple paths (rows, with letters corresponding to Figure 3.4). The significance (p values) of individual paths is assessed with Z-scores. Instead of damselflies, an obligate predator variable was used which represents the combined relative abundance of damselflies (M. modesta) and diving beetles (Copelatus sp.). Atlantic slope C. erethyzonfer Polypedilum spp. Scirtid spp. path Z p Z p Z p a: obligate predator ~ elevation -4.656 < 0.001 -4.656 < 0.001 -4.656 < 0.001 b: focal species ~ obligate predator -0.282 0.778 -1.599 0.110 0.238 0.812 c: Trentepohlia spp.~ obligate predator 0.144 0.886 0.144 0.886 0.144 0.886 d: focal species ~ Trentepohlia spp. -2.205 0.027 -3.609 < 0.001 -0.405 0.685 e: Trentepohlia spp.~ elevation 0.196 0.845 0.196 0.845 0.196 0.845 f: focal species ~ elevation 2.844 0.004 -4.710 < 0.001 3.476 0.001 ab: indirect effect mediated by obligate predator 0.281 0.778 1.512 0.130 -0.237 0.812 ed: indirect effect mediated by Trentepohlia spp. -0.195 0.845 -0.196 0.845 -0.177 0.860 acd: indirect effect mediated by obligate predator +          Trentepohlia spp. 0.143 0.886 0.143 0.886 0.135 0.892   147   APPENDIX C   C SUPPLEMENTARY INFORMATION TO CHAPTER 47                                                               7 Previously published as Online Resource 1 of Amundrud, S. L., Clay-Smith, S. A., Flynn, B. L., Higgins, K. E., Reich, M. S., Wiens, R. H. W., and Srivastava, D. S. (2019) Drought alters the trophic role of an opportunistic generalist in an aquatic ecosystem. Oecologia 189: 733-744  148   C.1  ANALYSES ON THE DATA FROM WHICH ALL CULEX SPP. WERE REMOVED FROM THE TOTAL PREY COUNT   Because the identity of the Culex species (Culex jenningsi vs. Culex erethyzonfer) was confounded with water treatment, we repeated the analyses on a subset of the data from which all Culex spp. were removed (i.e., mosquitoes are represented by Wyeomyia spp. only). The results for total prey abundance were qualitatively similar, suggesting that the confounding of Culex spp. identity with water treatment did not bias the results (Table C.1).   149     Table C. 1: Effects of tipulids and water on mosquitoes and chironomids in Experiment 1. Analyses were performed on a subset of the data from which all Culex spp. were removed, thus mosquitoes are represented by Wyeomyia spp. only. (a) Effect tests based on 2-way ANOVA, as summarized by F ratios and P-values. “Tipulid” represents three levels of tipulid treatments (no tipulid, caged tipulid, and uncaged tipulid), “Water” represents two water treatments (water present and absent), and “Tipulid x Water” represents their interaction. Significant P-values after sequential Bonferroni corrections (α1 = 0.05, α2 = 0.025, and α3 = 0.017, α’s are sequentially compared from largest to smallest P -value in each column) are indicated in bold. (b) Partitioning of tipulid effects into consumptive and non-consumptive effects. “Consumptive” effects compare caged vs. uncaged tipulid treatments for “Water present” and “Water absent” treatments, respectively. “Non-consumptive” effects compare no vs. caged tipulid treatments for “Water present” and “Water absent” treatments, respectively. Significant P-values after sequential Bonferroni corrections (α1 = 0.025, α2 = 0.013, and α3 = 0.008; α’s are sequentially compared from largest to smallest P -value in each column) are indicated in bold. Because of non-orthogonality, α’s are halved.  (a) 2-way ANOVA Tipulid Water Tipulid x Water Response F2,36 p F1,36 p F2,36 p Total prey abundance 14 < 0.001 58 < 0.001 9.2 < 0.001 Wyeomyia abundance 3.8 0.031 106 < 0.001 1.2 0.32   (b)  Tipulid effects Consumptive  Non-consumptive   Water present Water absent Water present Water absent Response t13 P t9 P t13 P t11 P Total prey abundance -0.65 0.52 6.3 < 0.001 3.4 0.005 0.02 0.99 Wyeomyia abundance 0.88 0.40 3.0 0.015 2.4 0.032 -0.31 0.76    150   C.2  EFFECTS OF WATER AND CAGE TREATMENT ON TIPULID GROWTH IN EXPERIMENT 1     Figure C. 1: Effects of water and cage treatment on tipulid growth. Tipulid growth was not affected by cage treatment (F1,17 = 4.10, p = 0.059), water treatment (F1,17 = 0.12, p = 0.73), or their interaction (F1,17 = 0.09, p = 0.77).    151   APPENDIX D   D SUPPLEMENTARY INFORMATION TO CHAPTER 58                                                               8 Previously published as Supporting Information of Amundrud, S. L., and Srivastava, D. S. (2019) Disentangling how climate change can affect an aquatic food web by combining multiple experimental approaches. Global Change Biology 25: 3528-3538  152   D.1  COMMUNITY COMPOSITION IN MICROCOSMS AND NATURAL BROMELIADS  Bromeliad-dwelling communities vary widely in species composition and relative abundances, and our goal was to replicate “typical” bromeliad communities in microcosms. We assembled communities consisting of the most common taxa in order to allow for replication, and show here that these artificially assembled communities resemble natural bromeliad communities occurring in a typical bromeliad leaf well.   D.1.1 Quantification of natural bromeliad communities We surveyed the food web inside bromeliads across the two study sites by quantifying the entire contents of 51 bromeliads (24 at MV, 27 at UGA) prior to the Experiment in 2016. The purpose of this survey was to (1) describe the bromeliad food web in the area, and (2) obtain insects to use in our experiment. A total of 42 bromeliads (21 at each site) were collected and hose rinsed in order to quantify their entire food web, while the contents of the remaining 9 bromeliads (3 at MV and 6 at UGA) were collected with a turkey baster. Both methods allow for a thorough quantification of the aquatic macroinvertebrate food web, but we were able to obtain the maximum water holding capacity, an index commonly used as bromeliad size, only for the rinsed bromeliads. For the basted bromeliads, the maximum capacity was estimated from a known correlation of bromeliad size and total aquatic invertebrate abundance at the two study sites (bromeliad capacity = 101.0567 + 1.1235*abundance, R2 = 0.28).  153   To quantify a “typical” community in a natural leaf well of maximum water holding capacity of 30 mL, we calculated the mean densities of all aquatic invertebrates at the two sites per 30 mL bromeliad capacity, and compare this to the artificially assembled communities inside our microcosms (Figure D.1). Although we found a total of 25 morphospecies between the two study sites, most were very rare and represented occasional occurrences. Of the 25 taxa, only nine occurred at densities higher than one individual per leaf well. Of these nine, only four occurred at densities of at least one individual at both sites (Figure D.1).   D.1.2 Microcosm communities represent natural bromeliad communities Microcosms were designed to mimic a bromeliad leaf well with a maximum capacity of 30 mL. The ambient (no drought) conditions in our microcosms thus mimicked a leaf well at a water level at 50% of the maximum capacity (i.e., 15 mL inside microcosms, or a water depth of about 2.1 cm), which is within the typical range of volume for one bromeliad leaf well (Amundrud and Srivastava 2015). Our goal was to replicate identical communities at the two sites, while keeping microcosm communities as close to natural communities as possible in both species composition and relative abundances. By assembling communities based on insects available to use from the bromeliad survey, we obtained communities that reflected those occurring naturally (Figure D.1: yellow diamonds). The artificially assembled communities represent communities for which the mean natural abundance of each species per 30 mL leaf well is at least one (Figure D.1, dashed line). Since we had to replicate communities exactly between the two sites, we only added species 154   whose abundance was greater than one at both sites (i.e., Scirtids and Oligochaetes were not included as they were rare at UGA, while Tanypodinae was not included as it was rare at MV).    Figure D. 1: Natural and experimental densities of taxa in our study. Red and blue points represent the mean abundances per 30 mL bromeliad maximum capacity from the survey (UGA: red, MV: blue). Error bars are standard deviations. Yellow diamonds represent abundances of taxa in our microcosms. The dashed line represents a mean abundance of one individual per leaf well, the threshold to include a given taxa in our microcosms. Note that the identity of the Culex species in the microcosms differs for the two sites, as C. erethyzonfer was common at MV and C. jenningsi was common at UGA.     155   D.2  EVIDENCE OF SUPPLEMENTARY PREY CONSUMPTION BY TIPULIDS  To test whether tipulid predation in our experiment was supplementary or compensatory to leaf processing, we examined the relationship of leaf loss to prey deaths. Unfortunately, we do not know what number of prey died from tipulid consumption vs. other causes like drought, so we performed a multiple linear regression where we predicted decomposition rates (as leaf loss) from prey mortality (as number of dead prey), tipulid presence or absence, and their interaction. To account for prey deaths from the drought treatment, we included drought length as a co-variate in the model.  Both tipulid presence and the interaction of prey mortality with tipulid presence significantly affected decomposition rates (Tipulid treatment: F1 = 18.8, p < 0.001; interaction of prey mortality with Tipulid treatment: F1 = 12.5, p = 0.001), while prey mortality and drought length had no significant effects on decomposition rates (R2 = 0.49). When tipulids were absent, prey mortality had negative effects on decomposition rates (Figure D.2), presumably because there were less detritivores (i.e., chironomids) left to consume detritus. When tipulids were present, decomposition rates increased with increasing prey mortality, suggesting that those tipulids that exhibit high predation rates are also consuming a lot of detritus. These results support a supplementary rather than a compensatory role of predation with decomposition by tipulids.    156        Figure D. 2: Leaf loss increases with prey mortality in communities with tipulids, but not in communities without tipulids.     157   D.3  TREATMENT EFFECTS ON PREY EMERGENCES  Prey emergence rates were not affected by the Origin→Location treatment or its interaction with tipulids or drought (Table D.1, Figure D.3), indicating that neither the space-for-time nor the common garden treatment affected prey emergence rates. However, the effects of drought on prey emergence rates depended on tipulid presence (Table D.1: T x D, Figure D.4). While prey emergence rates decreased as drought length increased in the absence of tipulids, emergence rates were generally low and independent of drought length in the presence of tipulids, likely because the adverse effects of tipulids on prey survival (see main document) resulted in fewer individuals available to emerge. Chironomids drove this result, as mosquito emergence rates during the experiment were insignificant (1.7 % of the mosquitoes and 14.1 % of the chironomids emerged over the timeframe of the experiment).    Table D. 1: ANOVA results for insect emergence rates (total prey and chironomids) from linear models. Model predictors are O→L = Origin→Location treatment (MV→MV, MV→UGA, or UGA→UGA), T = Tipulid treatment (present or absent), and D = Drought length. Because of few mosquito emergences, an analysis on mosquitoes was not possible.                    O→L T D O→L x T O→L x D T x D Emergences F2 P F1 P F1 P F2 P F2 P F1 P Total Prey 2.71 0.074 21.03 <0.001 11.32 0.001 0.09 0.916 1.18 0.314 5.86 0.018 Chironomids 2.88 0.064 22.30 <0.001 4.77 0.033 0.09 0.911 0.16 0.854 7.13 0.010 Mosquitoes -- -- -- -- -- -- -- -- -- -- -- --  158    Figure D. 3:  Effects of drought length, Origin→Location (O→L) treatment, and tipulid presence on the emergence rates of (a) total prey, (b) chironimids, and (c) mosquitoes. Experimental location represents current (MV) and future (UGA) climate. O→L treatments are indicated by color. Blue: MV→MV, green: MV→UGA, red: UGA→UGA. Tipulid treatment is indicated by solid circles and lines (tipulid present) versus empty circles and dashed lines (tipulid absent).   159       Figure D. 4: Partial residuals showing effect of tipulid treatment on (a-c) total prey emergences and (d-f) chironomid emergences for each community treatment. Blue: MV→MV, green: MV→UGA, red: UGA→UGA. Tipulid treatment is indicated by solid circles and lines (tipulid present) versus empty circles and dashed lines (tipulid absent).     160   D.4  HEAT TOLERANCE (CTMAX) OF TIPULIDS ORIGINATING FROM MV AND UGA  We experimentally determined the heat tolerance of tipulids originating from MV (n = 10) and UGA (n = 10) as their critical thermal maximum (CTmax). Tipulids were collected from bromeliads near MV and UGA between September and November 2015, and individually placed in a temperature controlled water bath that was heated at a rate of ~ 0.27 °C per minute. When tipulids ceased to move in response to a tactile stimulus, the temperature was recorded as CTmax.  The mean CTmax of tipulids originating from MV (40.75 °C ± 0.50 °C SE) was ~ 3 °C lower than the mean CTmax of tipulids originating from UGA (43.50 °C ± 0.60 °C SE; F = 12.46, p = 0.002; Figure D.5), which corresponds to the actual temperature difference between the two sites.     161        Figure D. 5: Heat tolerance (CTmax) of tipulids originating from MV (n = 10) and UGA (n = 10).   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0392690/manifest

Comment

Related Items