Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Complex spider webs as habitat patches : environmental filtering drives species composition Fernandez Fournier, Philippe 2016

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

Item Metadata


24-ubc_2016_september_fernandezfournier_philippe.pdf [ 1.22MB ]
JSON: 24-1.0308736.json
JSON-LD: 24-1.0308736-ld.json
RDF/XML (Pretty): 24-1.0308736-rdf.xml
RDF/JSON: 24-1.0308736-rdf.json
Turtle: 24-1.0308736-turtle.txt
N-Triples: 24-1.0308736-rdf-ntriples.txt
Original Record: 24-1.0308736-source.json
Full Text

Full Text

Complex spider webs as habitat patches: environmental filtering drives species composition  by  Philippe Fernandez Fournier B.Sc., McGill University, 2012    A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Zoology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)       August 2016  © Philippe Fernandez-Fournier, 2016  ii  Abstract Metacommunity theory has advanced understanding of mechanisms shaping community structure. Four main models (neutral, patch-dynamics, species-sorting, and mass-effects) have been recognized to explain these mechanisms, differing in their assumptions about the effects of environmental filtering and species traits on community composition. Here, I focus on complex, three-dimensional spider webs of two social and two solitary species as habitat patches for associated arthropods in a tropical rainforest in Ecuador. I used variance partitioning and various analyses of metacommunity structure to study the role of environmental filtering and dispersal in this system. I found that local patch characteristics, such as patch size and host species, predominantly affected local community composition. Webs of social spider species had higher richness, more variable communities, and proportionally more aggressive (i.e. predatory) web associates. Behavioral characteristics of the host spiders, such as sociality and aggressiveness, seemed to play an important role, as well, in shaping community composition on these patches. In a colonization experiment, there was indication of high dispersal rates at a short temporal scale and some evidence of species dominance at a longer temporal scale. I conclude that environmental filtering is responsible for the patterns of species distribution and that, given the conjunctive high dispersal and species specialization, the metacommunity patterns in this system seem to best be explained by a combination of the species sorting and mass effects models.    iii  Preface Philippe Fernandez-Fournier and Leticia Avilés designed this research project. Philippe Fernandez-Fournier collected the field data with the assistance of Sarah C. Bird (2014) and Samantha Straus (2015) with additional help from Mark Robertson and Esteban Calvache. Analyses were done by Philippe Fernandez-Fournier with supervision from Leticia Avilés.    iv  Table of contents Abstract ..................................................................................................................................................................... ii Preface ...................................................................................................................................................................... iii Table of contents .................................................................................................................................................. iv List of tables ............................................................................................................................................................ v List of figures ......................................................................................................................................................... vi Acknowledgements ............................................................................................................................................ vii Dedication ............................................................................................................................................................ viii Chapter One: Introduction................................................................................................................................. 1 Chapter Two: Environmental filtering drives associate species composition of complex spider webs ............................................................................................................................................................. 8 2.1 Synopsis .......................................................................................................................................... 8 2.2 Methods ....................................................................................................................................... 10 2.3 Results .......................................................................................................................................... 19 2.4 Discussion ................................................................................................................................... 30 Chapter Three: Conclusion ............................................................................................................................. 37 Bibliography ......................................................................................................................................................... 40 Appendices ........................................................................................................................................................... 45     v  List of tables Table 1: Predictions for questions addressed in this study under each metacommunity model. ........................................................................................................................................................................ 7 Table 2: Results of the partial redundancy analysis (RDA). .............................................................. 21     vi  List of figures Figure 1: Schematic representation of the four metacommunity models. ..................................... 3 Figure 2: Partitioning of the variation in community composition according to environmental and spatial explanatory variables using a PCNM analysis.. ................................ 20 Figure 3: Biplot of first two axes of the redundancy analysis constrained by Host and Web Size. .......................................................................................................................................................................... 22 Figure 4: Abundances of species that significantly drive the compositional difference in webs of the four host spider species.. ........................................................................................................ 23 Figure 5: Accumulation curves of species associated with webs of each of the four host spider species. ..................................................................................................................................................... 24 Figure 6: Relationship of (a) abundance, (b) richness, (c) density of associates with the size of the web they occupy and (d) results of the Tukey HSD on the intercept of each host species (no interaction terms). ..................................................................................................................... 27 Figure 7: Colonization by associates of newly-built A. eximius webs during a 24-day period.. ................................................................................................................................................................................... 28 Figure 8: Mean proportion of associate species of webs of A. eximius in new (24-days old) vs. old forest webs. ............................................................................................................................................ 29      vii  Acknowledgements This research was funded by NSERC Discovery (Canada) and James S. McDonnell foundation (USA) grants to Leticia Avilés. The author would like to thank his advisor Leticia Avilés for her insightful feedback and continued support. Catherine Hoffman, Mark Robertson, Ruth Sharpe, Luis Camacho and Suzana Diniz for their comments on the project. Sarah C. Bird, Samantha Straus and Esteban Calvache for invaluable help in collecting data. Greg Crutsinger, Mary O’Connor, Wayne Maddison and Diane Srivastava for comments on the project design and manuscript. Ingi Agnarsson and Nadine Dupérré for help on identification of associate species. The Jatun Sacha Biological Station and Ministerio del Ambiente, Napo, for providing the facilities, study area and permits for performing research in Ecuador. And finally Brittany Clark, for helping me maintaining a healthy work/life balance and for her perpetual support.      viii  Dedication        I would like to dedicate this work to my mother, who sparked my curiosity for the living world around me and supported me in harnessing it into a career.    1  Chapter One: Introduction Understanding the mechanisms underlying patterns of species distributions and interactions is a key component of community ecology, a field that currently is in the process of integrating local with regional scales. At a local scale, interactions amongst competing species is traditionally thought to decrease diversity through competitive exclusion, with coexistence resulting from niche differentiation (Chesson 2000).  Alternatively, local species composition may reflect transient dynamics resulting from differences among species in their colonization and competition ability (Tilman 1994). At the regional scale, on the other hand, the theory of island biogeography (MacArthur & Wilson 1967) considers local diversity a result of regional processes such as immigration and extinction. In trying to reconcile this dichotomy of patterns at different scales, the metacommunity concept looks at how interactions between spatially discrete communities at a regional scale affect the composition and dynamics of their communities at a local scale (Amarasekare 2003; Leibold et al. 2004). Whereas many studies have focused on elaborating theoretical frameworks, an increasing number of empirical studies are currently testing assumptions associated with these paradigms (Cottenie et al. 2003; Vanschoenwinkel et al. 2007; Guélat et al. 2008; Logue et al. 2011; Jimenez-Alfaro et al. 2015; Zheng et al. 2015). Along with addressing questions of scale, the metacommunity concept has also been useful in understanding how organisms are distributed according to environmental characteristics of the habitat patches (Holyoak et al. 2005). Two interrelated and complementary approaches to investigating the latter have been suggested: one focusing  2  on pattern, the other on mechanisms. The pattern-based approach tries to identify a structure of species distributions based on environmental conditions of the habitat patches (Leibold & Mikkelson 2002), whereas the mechanistic approach aims at identifying the processes behind species distributions (Leibold et al. 2004). To date, research has mostly focused on one or the other, with little integration of the two (Presley et al. 2010, Meynard et al. 2013). The pattern-based approach can help determine whether species, as a group, respond to the same environmental characteristics of habitat patches. To do so, one needs to examine if the species distribution, ordered along a common axis of local environmental characteristics, forms a coherent structure (Leibold & Mikkelson 2002; Presley et al. 2010). If species respond to different factors, they would fail to exhibit coherence along the same environmental characteristics and would have a distribution pattern that would not be significantly different from random. Although this approach allows one to identify the observed structure of a metacommunity, information about the underlying mechanisms responsible for such structure is lacking.  Within the mechanistic framework, four models have been put forward to explain species distributions in metacommunities (Fig. 1) (paradigms of Leibold et al. 2004). The neutral (NE) model (Hubbell 2001) implies that the patches are relatively homogenous in environmental conditions (environmental homogeneity) and that species are more or less equivalent in their life history, fitness and niche characteristics. Community assembly in this case depends entirely on extrinsic regional processes such as immigration and dispersal. Random extinctions eventually decrease diversity if these regional processes are absent or occur at low frequency. However, some communities can maintain their local  3  diversity at equilibrium with appropriate levels of immigration or speciation. Whether diversity increases within or among sites therefore depends on the relative strengths of extinction, speciation and dispersal.     Figure 1: Schematic representation of the four metacommunity models for two species, A and B. Figure from Falke & Fausch (2010) – modified from Leibold et al (2004). A species is dominant if it has a larger letter than another species on the same patch. Shapes that match for species and habitat patches represent adaptation of species for specific niches. Heavy, solid arrows represent rapid dispersal among habitat patches; light, solid arrows represent moderate dispersal; dashed arrows represent slow dispersal.   4  The patch dynamics (PD) model (Levins & Culver 1971; Tilman 1994) is similar to the NE model in its assumption that patches have similar ecological conditions (environmental homogeneity) and that species losses due to extinction or competitive exclusion are compensated by immigration. It differs from the NE model by considering that regional coexistence is the result of life-history differences among species. In most cases, this difference is shown through a trade-off between local competitive ability and dispersal, given that new patches form frequently enough to sustain species that are better colonizers than competitors.   Two further models assume that patches differ in local environmental conditions (environmental heterogeneity) and assume the presence of environmental filtering (specialization of species on patches with different properties). In the species sorting (SS) model (Levins 1969; Leibold 1998), communities are more limited by dispersal, meaning that local processes occur on a shorter time scale than colonization or extinction. Species that are better adapted locally thus have ample time to outcompete other species. In this model, because local coexistence is limited, species distributions closely match patch environmental conditions (Cottenie 2005; Chase et al. 2005). The mass effects (ME) model (Levin 1974; Mouquet & Loreau 2002), while also assuming environmental filtering, states that strong regional processes (emigration/immigration) preclude competitive exclusion. Dispersal, in this case, is fast enough in all species to substantially affect local dynamics and promote their co-occurrence on patches.  Therefore, the ME and SS differ from NE and PD by assuming environmental filtering; PD differs from NE by assuming a colonization-competition trade-off; and ME differs from SS by assuming that strong regional processes, such as dispersal, counteract  5  local dynamics. Although real metacommunities do not conform perfectly to a single identified model and might even be explain by a combination of two models (Cottenie 2005; Winegardner et al. 2012), this classification constitutes a useful framework to better understand increasingly complex metacommunity mechanisms working across multiple scales of space and time (Leibold et al. 2004). Establishing if species distributions are mostly explained by spatial dynamics or the environmental characteristics of the local habitat patches is challenging because local characteristics may be spatially correlated, making it hard to assign the observed pattern to spatial dynamics or environmental characteristics alone (Legendre 1993; Smith & Lundholm 2010). Variation partitioning is an effective tool to help distinguish between these two processes (Borcard et al. 1992; Legendre et al. 2005; Peres-Neto et al. 2006; Legendre & Legendre 2012). Under the niche models (SS and ME), local environmental variables will explain community structure better than spatial arrangement (e.g. geographical distance). Opposite patterns apply to the NE and PD models, where spatial arrangement will predominate in explaining community structure (McGill et al. 2006).    The challenge of assessing the effect of these environmental characteristics and identifying which of the four models best explains species distributions among metacommunities can be met with the use of natural microcosms as their small scale enables unambiguous measurements of boundaries, precise counts of species composition and high replication, all major impediments in studies of larger biological communities (Srivastava et al. 2004; Yamaura et al. 2016). Furthermore, these microcosms are not entirely closed systems due to processes of colonization and dispersal, which makes them  6  valuable for answering questions about metacommunity dynamics (Srivastava 2006; Agnarsson 2011; Petermann et al. 2015).  Here, I use as habitat patches the large, three-dimensional webs built by two solitary and two social spider species in a tropical rainforest in Ecuador (Appendix A). These webs provide habitat for foreign arthropod communities, which include other spiders, hemipterans, ants and coleopterans (hereafter called “associates”), which cohabit with the host spider species (Appendix B). Structural and physical features of the webs, as well as the spider species building the webs, were considered “environmental” characteristics of the patches. The webs of the four host spider species are similar in being tri-dimensional structures consisting of a finely meshed and non-sticky basal web, or “sheet”, topped with a tangle of interception prey capture lines above (Appendix A).  Social and solitary host species differ from each other not only in number of individuals building the webs (e.g. many small individuals vs. one large individual), but also in behaviors such as level of sociality and aggressiveness, which may affect the structure of communities living in these webs as well. For example, some hosts might be less aggressive towards web invaders than others. In addition to recording and characterizing the associates found in the tangle section of the webs, I also performed an experiment on one of the social species in which I let the spiders create new webs and monitored their colonization by associates. To establish if there is environmental filtering in this system and to determine which of the four metacommunity models best explains the observed patterns of species distribution, I addressed the following questions: (1) Were the communities predominantly affected by the local characteristics of the patches (as opposed to their spatial arrangement), as expected in the SS and ME models? (2) If the former were true (i.e.  7  species were host specific), what characteristics of the patches did the species align with? (3) Was there a temporal succession of species, potentially reflecting a colonization-competition trade-off as expected under the PD model, or was species dominance primarily due to sorting based on local environmental conditions, as expected under the ME and SS models? In addressing each question, I also attempted to characterize the extent to which the patches might be dispersal limited, with dispersal expected to be high in ME, low in SS, moderate in NE, and based on life-history tradeoffs in PD. Predictions under each metacommunity model are presented in Table 1. Table 1: Predictions for questions addressed in this study under each metacommunity model.  Q1. Variables affecting species distribution Q2. Species specialization Q3. Temporal succession Species dispersal Patch dynamics (PD) Spatial None Yes Differential* Neutral model (NE) Spatial None No Moderate Mass effects (ME) Environmental High No High Species sorting (SS) Environmental High No Low * Species have different dispersal abilities, as a result of a colonization-competition trade-off.      8  Chapter Two: Environmental filtering drives associate species composition of complex spider webs 2.1 Synopsis I used the large, three-dimensional webs of four spider species – two solitary and two social – to explore the processes that may be responsible for the assemblage of local communities in a metacommunity.  These webs are typically colonized by a variety of organisms.  Four models have been recognized to explain the mechanisms behind species compositions in metacommunities, differing in their assumptions about the effects of environmental filtering and species traits on community composition. The neutral (NE) and the patch dynamics (PD) models, contrary to the mass effects (ME) and the species sorting (SS), assume no effect of environmental filtering. PD differs from NE by assuming a colonization-competition trade-off, whereas ME differs from SS by assuming that strong regional processes, such as dispersal, counteract local dynamics. To establish whether there is environmental filtering in this system and to determine which of the four metacommunity models best explains the observed patterns in associated arthropod communities of complex, three-dimensional spider webs, I addressed the following questions: (1) Were the communities predominantly affected by the local characteristics of the patches (as opposed to their spatial arrangement), as expected in the SS and ME models? (2) If the former were true (i.e. species were host specific), what characteristics of the patches did the species align with? (3) Was there a temporal succession of species, potentially reflecting a colonization-competition trade-off as expected under the PD model, or was species dominance primarily due to sorting based on local environmental conditions, as expected under the ME and SS models? In addressing each question, I also  9  attempted to characterize the extent to which the patches might be dispersal limited, with dispersal expected to be high in ME, low in SS, moderate in NE, and based on life-history tradeoffs in PD. I used large, three-dimensional webs built by two solitary and two social spider species in a tropical rainforest in Ecuador as habitat patches (Appendix A). Structural and physical features of the webs, as well as the spider species building the webs, were considered local “environmental” characteristics of the patches. Spatial arrangement of the webs was also recorded in the form of GPS coordinates. In addition to recording and characterizing the associates found in the tangle section of the webs, I also performed an experiment on one of the social species (A. eximius) in which I let the spiders create new webs and monitored their colonization by associates. I found that patch characteristics, such as size and host species as well as social system, predominantly affected local community composition. Webs of social spider species had higher richness, more variable communities, and proportionally more aggressive (i.e. predatory) web associates. In the colonization experiment, there was no apparent pattern of temporal succession at a short temporal scale, but some evidence of species dominance at a longer temporal scale. I concluded that the species distribution patterns in this system can best be explained by a combination of species sorting and mass effects, but that the four established metacommunity models are potentially not mutually exclusive from one another in explaining mechanisms of species distributions in metacommunities.   10  2.2 Methods Study species One of the solitary host species from which web associates were collected, Kapogea sexnotata (Simon, 1895), is an orb-weaver (family Araneidae) which, unlike most araneids, builds a horizontal and finely meshed platform surrounded by a tangle of irregular webbing (Levi 1997). The other solitary host spider species, Aglaoctenus castaneus (Mello-Leitão, 1942), is a funnel-weaving wolf spider (family Lycosidae), which builds a flat, slightly basket-shaped sheet with a funnel retreat on one side. This genus, although containing only 5 species (Platnick 2016), is common in South America and is one of the only New World genera of Lycosidae reported to build sheet webs with funnel-shaped retreats (Santos & Brescovit 2001). The two social species, Anelosimus eximius (Keyserling, 1884) and Anelosimus domingo (Levi, 1963), belong to the family Theridiidae and form colonies that may contain from a few to thousands of individuals who cooperate in prey capture, feeding and brood care (Avilés 1997).  In both species the colonies are housed in webs consisting of a basket-shaped and finely-meshed basal sheet topped by a relatively large tangle of prey capture lines (Appendix A).  A. domingo spiders, however, are about one third the size of A. eximius and have been shown to be individually faster to respond to struggling prey (Guevara & Avilés 2011).   As shown in the Appendix (Appendix B), the species associated with the webs were categorized as being aggressive or passive based on their observed behavior and known life history strategies (e.g. kleptoparasitic, predatory or scavenger). Passive associates included some kleptoparasites, scavengers and other commensals who would scavenge  11  small prey ignored by the host (Koh & Li 2003) or feed on large prey items captured simultaneously with the host (Vollrath 1979). Aggressive associates were observed actively stealing prey from the host or preying on the host itself and included some kleptoparasitic and predatory species.  Field sampling All associates were collected from webs occurring in lowland tropical rainforest habitat at the Jatun Sacha Biological Reserve, near Tena, Ecuador (S 1.07° W, 77.61°, Napo Province) in the summers (June- August) of 2014 and 2015. The number of the various associate species per web were recorded for a total of 150 webs (66 in 2014; 85 in 2015) distributed over a 2.2 × 2.6 km area (Appendix C). Individuals were directly collected from the webs and brought to a laboratory facility where they were identified based on morphology and life history (Ubick et al. 2005) using a stereomicroscope (AmScope SM-1TSW2, ToupView software) with a 5MP digital camera. Further identification was made using identification guides and descriptions specific to families. Species that were observed less than 10 times throughout the data collection period were not included in the dataset.  To determine whether the species encountered in webs were also present in the surrounding environment, I recorded the presence of any recognizable web associates along 10 × 2 × 2-meter transects (in 2015) established at 10 randomly assigned locations in the study area. I also carried out a colonization experiment using the social A. eximius.  I collected entire nests and then allowed the spiders to build a new web (i.e. patch), which I monitored for colonization by associates every two to four days for 24 days. In this case, associates  12  could not be brought to the laboratory for precise identification and therefore the resolution was decreased in this experiment and some species were lumped together because they were indistinguishable in the field.  All analyses were performed using the R software version 3.1.2 (R Development Core Team 2013).  Variation partitioning  To first assess if characteristics of patches were predominant in shaping community structure in this system, I partitioned the variation of species abundance data into four independent explanatory components: pure environmental, pure spatial, spatial component of environmental influence, and undetermined. The environmental variables, chosen based on a forward model selection, were: host spider species and size of the web. The host spider species was considered as environmental variables since webs of the different species differ in their architecture, quality and silk thickness (see Appendix A). The size of the webs was calculated as the total volume of the tangle comprised by the capture lines above the basal sheet platform (and also under it in the case of Kapogea), which has typically a conical shape in A. domingo, A. eximius and Aglaoctenus sp. and an ellipsoid shape in Kapogea.   The spatial variables consisted of a matrix of two-dimensional geographical coordinates, longitude (x) and latitude (y), completed by producing orthogonal (linearly independent) spatial variables using an analysis of principal coordinates of neighbor matrices (PCNM). This procedure ensures that complex patterns over a much wider range of spatial scales can be extracted from the species distribution and abundance data and  13  allows the modelling of any type of spatial structures (Borcard & Legendre 2002). The PCNM method is part of a family of methods called MEM (Moran’s eigenvector maps; Dray et al. 2006). Only webs collected in the year 2015 were considered in this analysis, as the number of webs collected in 2014 did not include enough webs of all four hosts. To avoid an overestimated explanation of variation, a forward model selection (based on adjusted R2 and P-value) of explanatory variables was applied on the resulting thirty-two (32) eigenvectors obtained from the PCNM analysis. The eigenvectors 4 and 21 were retained.  Variation partitioning is a good way to partial out the spatial explanatory component of species abundances to determine the effect of patch characteristics vs. spatial arrangement in structuring communities. In situations where patch characteristics (biotic and abiotic) determine most of the variation of the communities, the relative amount of variation explained solely by patch characteristics would be high. The proportion of variation attributed to characteristics of the patches may also depend on the specific spatial configuration of the environmental variables, which would be reflected in the covariance of environmental and spatial variables, or “spatial autocorrelation” (Legendre 1993).  Finally, some variation may depend solely on the spatial arrangement of the patches, reflecting whether species are dispersal limited or not. I used a constrained ordination to measure the amount of variation in species abundances across patches that can be explained by a set of environmental and spatial variables, as recommended by Legendre et al. (2005). The type of constrained ordination used was a redundancy analysis (RDA) on Hellinger-transformed species abundances. Transformation gives more equitable weight to all species by reducing the influence of the most abundant species, and placing  14  emphasis on differences in proportional representation of species between webs rather than their absolute abundance differences (Legendre & Legendre 2012). There was a high number of zero abundances in my species data set and high variation in abundances, with various rare species that do not necessarily represent the communities. Given this, RDA is preferred over a Canonical Correspondence Analysis (CCA) because it gives less weight to rare species and because abundances are expected to have a linear relationship with the variables, as opposed to a unimodal relationship, which is an assumption in CCA. All analyses of variation partitioning were done using the package vegan (Oksanen et al. 2016). Species specialization To detect differences in species composition among patches and to analyze the extent of environmental filtering, I performed a partial RDA with environmental characteristics (host species and web size) as explanatory variables, controlling for collection year (2014 or 2015). Eigenvalues of each patch along the first two axes of the RDA (representing most of the variation) were retained for further analyses. Forward selection of both environmental and spatial explanatory variables led to the retention of host and web size only as the main variables affecting species composition. Interactions between explanatory variables were not analyzed because they did not significantly improve the explanatory power of the model in the forward selection procedure. The community data matrices were Hellinger-transformed, thereby giving lower weights to rare species (Legendre & Gallagher 2001). A permutational multivariate analysis of variance (PERMANOVA) of Bray-Curtis distances between communities with 999  15  permutations was used to assess statistical significance of the effect of environmental variables on community composition (package vegan; Oksanen et al. 2016). As an indication of species differentiation, I measured which species of associates were driving the compositional differences between patches using a generalized linear model (with the same explanatory variables as in the RDA) with 999 permutations and negative binomial errors (selected based on residual deviance) comparing the abundance of each associate species in the four host web types (package mvabund; Wang et al. 2012).   To assess the accumulated richness of associates on each host species, species accumulation curves were obtained by sampling from the species abundance data and finding the expected averaged rarefied richness at any given number of webs sampled (package vegan; Oksanen et al. 2016).    Social and solitary host species differ from each other in behavior (e.g. aggressiveness towards web invaders), as well as number of individuals in the webs (e.g. many small individuals vs. one large individual), which may affect the structure of communities living in these webs. I therefore tested for differences in variability (i.e. “indefiniteness”) of species composition between social (A. domingo and A. eximius) and solitary (Kapogea and Aglaoctenus) hosts, as measured by the multivariate dispersion of communities of given hosts from their host centroid in multivariate space (package vegan; Oksanen et al. 2016). Groups of communities with more variable species compositions have higher dispersion.  I also compared the proportion of “aggressive” associates (active kleptoparasites and predators) in social and solitary host species using a linear mixed  16  effects model, as I expected that predatory associates might be more common in social species, where the hosts themselves could be prey. I further tested if the distribution of species among patches had a coherent structure according to a common axis of environmental variables, which gives an estimate of the extent to which different associate species may respond to the same characteristics of the patches on which they co-occur. If species as a whole do not respond to the same characteristics, their distributions will not show a pattern of coherence, indicating that species are not distributed significantly different than expected by chance when patches are ordered according to common ordinal axes (Leibold & Mikkelson 2002). This is the first step in analyzing if there is a pattern in the associate species distribution, which consists of ordering the web communities in a site-by-species matrix according to a regular canonical ordination analysis and identifying one or more axes (corresponding to the environmental variables) responsible for the structure among communities. This ordered site-by-species incidence matrix is then compared with random expectations through randomization, which allows coherence to be characterized. If coherence is significant, further steps have to be taken to identify the exact pattern (i.e. turnover and/or nestedness). Coherence was estimated by comparing the number of observed embedded absences in the distribution of species across patches with the number expected under a null model (package metacom; Dallas 2015). An embedded absence is identified in the ordered species occurrence matrix as an absence that is bounded on either side by a presence (Leibold & Mikkelson 2002; Presley & Willig 2010). The order in which coherence was measured where the eigenvectors of each patch along the first two axes of the RDA. The coherence observed was  17  then compared to a null model that maintained the species richness of patches (row totals) and filled species ranges (columns) based on their marginal probabilities (999 iterations).  If variation in species compositions was significantly explained by their host species, I examined if there was a pattern of nestedness of communities within each host according to the size of the webs. A nestedness pattern would imply that communities on smaller patches are a subset of those on larger ones (Ulrich et al. 2009). I performed this analysis on species occurrences (presence/absence) and used a nestedness metric based on overlap and decreasing fill (NODF) as a measure of level of nestedness (Almeida-Neto et al. 2008). The null model used conserves the dimensions and fill (i.e. presences) of the original species occurrence matrix but randomly assigns occurrence values (null model SS: Beckett et al. 2014; Staniczenko et al. 2013). Although potentially preferred for species co-occurrence analyses, the most conservative model (fixed row and column totals) cannot be applied to small matrices such as the one in this study as there are not enough possible rearrangements (Ulrich & Gotelli 2007).  Finally, to better understand how patch size affected community composition across the various host species, I analyzed the relationship between abundance, richness and density of associates with the size of the webs of the various host species using a linear model.  Temporal succession To assess patterns of colonization and exclusion over time I performed a colonization experiment with newly built webs of A. eximius and monitored the colonization by associates every two to four days for a total of 24 days. Species or groups of  18  species were counted at every survey date and the entire web was collected at the end of the experiment to have a more accurate assessment of species composition. A pattern of interspecific temporal succession on webs would imply that species differ in their colonization and competitive abilities due to a colonization-competition trade-off and would be consistent to the PD model (Tilman 1994).  I investigated the short-term succession within 24 days, as well as long-term succession by comparing 24-day old webs with webs naturally occurring in the forest.     19  2.3 Results Species associated with webs A total of 22 species were identified as common or obligate nest associates of the four host spider species (Appendix B).  Nine of the 22 nest associates were spiders belonging to the subfamily Argyrodinae (Araneae: Theridiidae). Two of them, Rhomphaea sp. and Neospintharus sp., were seen as predatory, along with a spider from the genus Mimetus (Araneae: Mimetidae) and an unidentified jumping spider from the subfamily Amycini (cf. Hypaeus). Scavenger species observed included two hemipterans, Ranzovius sp. and an unidentified emesine reduviid species (subfamily Emesinae), and one coleopteran in the genus Paratenetus, which were mainly seen feeding on unwanted or already eaten prey of the host spider. Three ant species were frequently observed in the webs and, when present, occurred in high numbers. One ant species, Crematogaster sp., was considered as an active kleptoparasite, as it was observed actively stealing prey and even building satellite nests inside the spider webs, whereas the other two ant species, Wasmannia sp. and Tetramorium sp., were seen mostly scavenging on discarded prey.  Ten out of the 22 web associates previously identified were encountered at least once along the 10 forest transects. All of these species, except for one, constituted less than 2% of all spiders and insects recorded along the transects. The exception was Araneidae sp.1, which constituted almost 10% of the species seen in the forest. Two spider species, Argyrodes sp.1 and Faiditus sp.3, were always found as associates of other webs.   20  Variation partitioning  When partitioning the variation in species composition across host webs, environmental characteristics of the patches explained approximately 31.5% of the variation, of which 0.7% constituted covariance between environmental and spatial variables (Fig. 2). Spatial arrangement of the webs, on the other hand, explained about 2.7% of the variation, with 2.0% of this variation due to the spatial arrangement alone. The total amount of explained variation was therefore 33.5%, resulting in an amount of unexplained (residual) variation of 66.5%. Twelve webs were removed from this analysis because of unavailable geographical coordinates.   Figure 2: Partitioning of the variation in community composition according to environmental and spatial explanatory variables using a PCNM analysis for the year 2015. Environmental variables include host species and web size. Spatial variables are the three PCNM eigenvectors selected in a forward model selection. The four components are proportions of variation explained (adjusted R-square) by the following variables: pure environmental, spatial component of environmental influence (spatial autocorrelation; overlap section), pure spatial, and undetermined (residuals).    21  Species specialization  Host species and web size both significantly affected community composition across patches, with host explaining a much larger proportion of variation in the redundancy analysis results (Table 2). Patches tended to differentiate along the first ordination axis according to level of sociality (social vs. solitary host species), whereas they tended to vary according to size of the web along the second axis (Fig. 3). Community composition varied significantly between webs built by different host species, with 10 out of the 22 associate species significantly driving the compositional difference (Fig. 4). The other 12 species did not show any significant pattern in their distribution and were seen in similar proportions across webs of different hosts. Overall, most of the associate species were recorded in all web types but in different proportions for the 10 species driving the compositional difference (Fig. 4). Table 2: Results of the partial RDA after controlling for Year of collection. Effects of Host spider species, volume of web (log-transformed) on community composition of web associates. Significant p-values are in bold.   Partial RDA  % of variation F P All associates Host 22.0 14.5 0.001  Patch volume 2.8 5.5 0.001  Year (covariate) 3.1 6.1 0.001  Total 27.1 12.0 0.001      22   Figure 3: Biplot of first two axes of the redundancy analysis constrained by Host and Web Size (ln-transformed). On the graph on the left, points represent individual web communities. Points (patches) that are closer to each other have a greater species composition similiarity. Red points: Aglaoctenus; purple points: Kapogea; green points: A. domingo; blue points: A. eximius. Plot on the right represents the species composition of associates, showing in red the species that significantly drive the compositional difference according to results from a multivariate generalized linear model performed for each associate species. Centroids of communities on webs of the four host species are represented by “+” symbols matching the host color.   23   Figure 4: Abundances of species that significantly drive the compositional difference in webs of the four host spider species. Darker bars indicate the host species for which each associate species has the highest mean abundance.   24  When comparing social (A. domingo and A. eximius) and solitary (Kapogea and Aglaoctenus) hosts, the social species had a significantly greater variability (multivariate dispersion) of species composition than the solitary species (ANOVA and a posteriori Tukey HSD, F3,146 = 23.9, p < 0.001). Social host species also had a significantly higher proportion of “aggressive” associates (active kleptoparasites and predators) compared to solitary ones (linear mixed effects model, χ2 = 13.9, df = 4, p < 0.001) (package lme4; Bates et al. 2014). The species accumulation curves showed that communities forming on A. domingo and A. eximius webs had higher accumulated species richness (Fig. 5), reaching 22 species for both species at the maximum number of webs sampled. The accumulated richness in webs of Kapogea and Aglaoctenus was 18 and 16 species, respectively.  Although richness on webs of A. domingo tended to accumulate more slowly compared to other host species, the accumulation trends of the curves of all host webs tended to level off. Webs of social host species therefore tended to have more variable communities, harbor a higher proportion of aggressive associates and host communities with higher richness.  Figure 5: Accumulation curves of species associated with webs of each of the four host spider species.   25     Overall species distribution did not show significant coherence along either of the first two axes of the RDA (i.e. explaining most of the variation in the analysis) (RDA axis 1: No. of embedded absences = 1874, simulated mean = 1886, SD = 83.0, p = 0.89; RDA axis 2: No. of embedded absences = 1874, simulated mean = 1894, SD = 79.9, p = 0.80). The site-by-species matrix ordered along these axes, representative of local patch characteristics, had therefore no significant coherence and was not significantly different than random expectations, making further analyses of pattern unnecessary according to the framework presented by Presley et al. (2010). When separately looking for patterns in communities of each host species, however, webs of all four spider species hosted communities significantly nested based on the size of their webs (Aglaoctenus: measured NODF:  48.5, mean of null distribution:  12.3, p = 0.001; Kapogea: measured NODF:  40.1, mean of null distribution:  16.8, p = 0.001; A. domingo: measured NODF:  31.6, mean of null distribution:  8.0, p =  0.001; A. eximius: measured NODF:  44.9, mean of null distribution:  13.6, p =  0.001). Nestedness was not significant in neither of the host webs when using a more conservative null model that conserved the same number of occurrences per row and columns. When analyzing the effect of web size alone, I found that abundance, richness and density of associates were all significantly influenced by web size. Regarding web size itself, social species had significantly greater variance in web size compared to the solitary ones (F test; F85, 63 = 2.39, p < 0.001). When looking at differences among host species of the same sociality, the social species differed from each other in web quality, with A. domingo having greater density of threads in the tangle (where associates occur) (Pearson, χ2 = 19.23, p < 0.001) and cleaner webs (Pearson, χ2 = 18.61, p < 0.001) than A. eximius. Web  26  quality was not taken for solitary host species. Among solitary species, the Kapogea species had webs significantly higher from the ground than webs of Aglaoctenus (t-ratio = 2.03, df = 37, p = 0.025). The abundance of the associates significantly increased with the size of the web for all hosts (Adj. R-squared = 0.70, F4,145 = 87.74, p <0.001, Fig. 6a). The richness of associates was also positively correlated with web size, although a smaller amount of the variance was explained by the model (Adj. R-squared = 0.37, F4,145 = 23.34, p <0.001, Fig. 6b). The density of associates (associates/cm3) significantly decreased with increasing size of the web for all hosts (Adj. R-squared = 0.76, F4,145 = 117.5, p <0.001, Fig. 6c).  The relationships of abundance, richness and density to size of the webs differed among host species. When performing Tukey HSD a posteriori tests for each model (without the interaction term with host species, which was non-significant), A. domingo had a significantly lower abundance of associates than the other hosts and A. eximius had significantly higher richness than A. domingo and Aglaoctenus (Fig. 6).  27    Figure 6: Relationship of (a) abundance, (b) richness and (c) density of associates (per cubic meter) with the size of the web they occupy. All variables are log10-transformed. Results of the Tukey HSD on the intercept of each host species (no interaction terms) are on the bottom right table with different letters stating significant difference in order of magnitude (A = highest, B = lowest). Temporal succession Species did not show temporal succession at a short scale (0-24 days, Fig. 7). All species tended to colonize the webs rapidly, with species colonizing a high proportion of webs (median proportion of patches occupied by associates = 0.92). Variance in mean ant abundance was large due to the fact that ants tended to be present in high abundance or not at all.   28   When assessing temporal succession at a longer time scale, the relative abundance of three species assumed to be specialized on webs of this host species (Faiditus sp. 1, Faiditus sp. 6 and Ranzovius; see Fig. 4) was significantly higher in older webs naturally encountered in the forest compared (n = 42) to 24-days old transplanted webs (n = 25) (Welch t-test; Faiditus sp. 1: t-ratio = 4.43, df = 49.1, p < 0.001; Faiditus sp. 6: t-ratio = 2.32, df = 54.3, p = 0.024; Ranzovius: t-ratio = 2.77, df = 47.4, p = 0.008; Fig. 8). On the other hand, Philoponella, ants, and the amycine salticid (cf. Hypaeus) all decreased in proportion in older webs compared to 24-days old webs, although not significantly so (Fig. 8). All other associate species did not show a significant pattern of change in proportion.   Figure 7: Colonization by associates of newly-built A. eximius webs during a 24-day period. Abundances are log10-transformed. Lines connect means ± standard errors of each species at each monitored day.   29   Figure 8: Mean proportion (arcsin-transformed with confidence intervals) of associate species of webs of A. eximius in new (24-days old, n = 25) vs. old forest webs (n = 42). Species colors vary from red (highest decrease in proportion) to green (highest increase in proportion).  Star symbols next to associate species indicate a significant difference in mean proportion between new and old webs.   30  2.4 Discussion Four models have been proposed to explain the composition of local habitat patches in a metacommunity setting.  In both the neutral (NE) and patch dynamics (PD) models it is assumed that species exhibit no specialization to patches and that patches are mostly homogenous, whereas environmental filtering is present in the species sorting (SS) and mass effects (ME) models.  Additionally, local species composition is expected to be affected by dispersal in and out of patches, which is assumed to be moderate in the NE model, based on life history trade-offs in the PD model, limited in the SS model, and relatively extensive in the ME model.  Overall, environmental filtering appeared to be driving the composition of associates of the three-dimensional spider webs studied. Therefore, species composition in this system could be explained by the SS model, given that (1) species were sorting according to local characteristics of patches; and (2) the relative abundance of specialized species was significantly higher in older webs compared to newer webs. However, species composition could also be explained by the ME model given that (1) although differences among communities depended largely on local characteristics of the patches, many associate species also occurred in all patch types; and (2) dispersal limitation was almost entirely absent and colonization rates (i.e. dispersal) of associate species were high. The composition of patches in this system could therefore be a case of species sorting with species having relatively high dispersal rates and broad tolerances to patches with suboptimal characteristics or a case of mass effects with dispersal not being high enough to entirely counteract local processes of species sorting.     31  When examining the phylogenetic composition of the associates, species belonging to the subfamily Argyrodinae were constituting a significant proportion of the identified web associates. Species in this subfamily are known for their association with a wide range of spider species (Agnarsson 2002) and are sometimes even considered as group-living associates specializing on large host webs (Su & Smith 2014). Among the predatory species, spiders in the genus Mimetus (Araneae: Mimetidae) are also frequently seen in association with social spider webs (Perkins et al. 2007; Pruitt & Riechert 2011), remaining in close proximity to the webs and venturing onto them to capture the resident using aggressive mimicry (Whitehouse 1987; Romero & Flórez 2014). The hemipteran and coleopteran scavenger species found in this study (Ranzovius sp., Emesinae reduviid and Paratenetus sp.) are commonly observed in spider webs and are considered as commensals (Wygodzinsky 1966; Henry 1984; Wheeler Jr & McCaffrey 1984). The kleptoparasitic species of ant (Crematogaster sp.) has previously been documented building “satellite” nests inside spider webs (Fowler et al. 2013), as in this study. These species may thus have the ability to be temporarily associated with webs of sub-optimal characteristics and therefore have better chances of colonizing webs to which they are better adapted. Additionally, although results from the transects did not show that species were abundant in the matrix, the forest could potentially be the source for some species, rather than other webs. Lastly, 15 out of the 22 associate species were recorded in all web types (plus 4 out of 22 recorded in 3 web types), suggesting high dispersal rates, but also some degree of specialization, given they occur in different proportions. Web characteristics explained over 10 times more the variation in species composition than spatial arrangement of the webs. Patch homogeneity is thus violated in  32  this system, suggesting the PD and NE models cannot adequately explain the observed patterns. However, the amount of unexplained variation, which is a combination of nondeterministic ecological fluctuations and potentially unmeasured deterministic (biotic and/or abiotic) variables, was fairly high (66.5%). One might argue that this result can be due to stochastic processes and has consequently connections to the NE model, which assumes that the population dynamics are primarily driven by ecological drift and dispersal and are not habitat dependent (Legendre et al. 2009). However, a high amount of unexplained variation is not unusual in similarly complex ecological systems (Cottenie et al. 2003; Cottenie 2005; Vanschoenwinkel et al. 2007; Alahuhta et al. 2014; Dallas & Presley 2014). It is also possible that the portion of explained variation would have been larger if other patch characteristics, such as quality of the webs or host species traits, had been measured. However, given the potentially complex interactions between associate species, and the availability of other habitable webs of spiders with different characteristics, I suspect that, unless all large, complex webs were to be systematically collected in this study area, the proportion of unexplained variation will remain high. This pattern is therefore most probably attributable to the complexity of the system rather than being evidence in favor of the NE model. The absence of a strong effect of spatial arrangement on community composition further suggests that dispersal is not limiting in this system, which is in greater agreement with the ME than the SS model. The host species constituted most of the explained variation in community composition. Each host spider, although building three-dimensional webs of roughly similar characteristics, unsurprisingly had differences in the structure and consistency of their webs as well as in their behaviors. These characteristics probably acted in concert to  33  give rise to the significant differences observed between communities in webs of different hosts. Some of these characteristics could qualitatively be determined from the first main axis of the RDA. Along this axis, for instance, patches seemed to differentiate in species composition according to the social system of their host spider (social or solitary). When looking at this ordination axis alone, Aglaoctenus and Kapogea were more similar to each other and were separated from the two social Anelosimus species, which were also more similar to each other. Along the second ordinal axis, patches tended to align based on the size of the webs, with the larger webs of Kapogea and A. eximius, on one end, (see Appendix A), and those of Aglaoctenus and A. domingo, on the other. When both axes are plotted on the biplot of Fig. 3, one can thus see that all four hosts seem to have diverging species compositions mostly based on host social system and web size. Additionally, social species had significantly greater variance in web size compared to the solitary ones, which might be responsible in part for the difference in variability of associated communities. Communities associated with the two social species further differed from those on the solitary ones in having higher accumulated richness and in harboring relatively more aggressive associates.  The former pattern could be explained by the fact that social species, are much smaller in body size relative to the associates than are the solitary species (see Appendix A), thus potentially providing an additional resource through an abundance of potential prey for associated predators.  This may explain why there were significantly more predatory associates in webs of A. eximius and A. domingo than in the solitary species. Social species, by having reduced intraspecific aggression (Agnarsson 2002), may also have greater tolerance to non-conspecifics compared to solitary species.  Reduced aggression toward web invaders could thus explain the higher accumulated richness and greater  34  variability in species composition in social compared to solitary hosts. This feature of social species might also provide further insight into a cost of living in permanent social groups. By developing higher peer tolerance to allow groups of individuals to cooperate and be fully social, species would also allow a greater diversity and abundance of web invaders. Additionally, the low number of host species sampled in this study and the fact that two of them (A. eximius and A. domingo) are not phylogenetically independent makes conclusions about social spiders more complicated. Given the fact that closely related species often exhibit phylogenetic conservatism and consequently have very similar phenotypes, the two social species are expected to have quite similar web types and associates. Therefore, although these two species are almost the only social species in this area, it is difficult to infer what associate communities of social spiders are as a whole from this one congeneric pair.  When considering the four host species combined to test for metacommunity structure, no pattern of coherence of the species distributions with the ordered characteristics of patches, coinciding with sociality and web size (first two RDA axes), was present. The dominance of host specific associates is likely to be what explains the lack of metacommunity structure associated with a single common axis of patch characteristics. The particular responses of species give rise to emergent structures on different, independent axes and an apparently random structure is formed (Leibold & Mikkelson 2002). When looking at structure of species distribution within host species, however, the results were different. The community of a given host had a significant pattern of nestedness linked to the size of the web they occupied, implying that communities on smaller patches were a subset of those on larger ones and that patch size was one of the  35  main characteristics influencing the structure of communities assembled in webs of each host (Reeves & Wright 1992). However, since some species were specialized to specific hosts, this nested pattern was made of different species across host types.  One might suspect that other characteristics specific to a certain host (or subset of hosts) are influencing the species composition at a finer scale. Thus, behavioral differences among host species might be partly responsible for differences in species composition. For instance, A. domingo has been found to be more aggressive than A. eximius because of its overall faster reaction time when handling prey and the involvement of a greater number of individuals in the capture of prey (Guevara & Avilés 2011). This higher level of aggressiveness might explain the lower abundance of associates in A. domingo compared to A. eximius webs. The species also differed in web quality, with A. domingo having greater density of threads in the tangle (where associates occur), which were also cleaner than those of A. eximius. Unique characteristics of webs built by the hosts could thus also further explain the differentiation in species composition of associates they harbor. Although the webs of Kapogea are similar to those of other species in this study in having a basal sheet with a tridimensional tangle above, their basal sheet is less dense than that of the other species (pers. obs., see Appendix A).  Kapogea’s webs are also built significantly higher from the ground than webs of Aglaoctenus.  These characteristics could provide further explanation to the difference in associate composition in webs of the two solitary species. The colonization pattern in the first 24 days in webs of A. eximius, albeit at a relatively short temporal scale, was not typical of species having a competition-colonization trade-off, such as in the PD model. At a longer temporal scale, on the other hand, proportions of most associate species tended to differ between new and old webs,  36  with a significant increase in proportion for three species. Because of the lack of temporal succession, whereby pairs of species would completely replace one another, the PD model is not adequate in explaining this pattern. The difference in proportional abundance of only specialized species suggests that species sorting is occurring but that mass effects might be present as well for more generalist species. Additionally, most species tended to colonize the webs rapidly, with a high proportion of webs being colonized in the first 24 days of a newly formed web. This pattern strongly indicates that associate species have overall high colonization rates and consequently non-trivial dispersal abilities. Under the ME model, one might hypothesize that dispersal is nevertheless not high enough to totally counteract local sorting of associate species. On the other hand, one might also hypothesize that specialized species are slightly more dispersal limited than more generalist species, which would explain their higher proportional abundance after a longer time.     37  Chapter Three: Conclusion Altogether, co-occurrence and temporal patterns in community composition of associates in large, complex three-dimensional spider webs in a tropical rainforest showed a strong effect of environmental filtering in the composition of their associates, with apparently little dispersal limitation.  However, given that many associate species also occurred in all patch types, the environmental filtering affecting species composition of patches in this system could reflect either a case of species sorting, with species having relatively high dispersal rates and species having fairly broad tolerances to patches with suboptimal characteristics, or through mass effects with dispersal not being high enough to counteract species sorting.  In this study, I used a simple model system to help understand the factors affecting species composition on habitat patches. Obligate web associates and their host webs form a useful system in which one can investigate the roles of physical characteristics of patches and other system-specific properties (such as host behavior, in the case of this study) in explaining the distribution and structure of species assemblages in habitat patches. However, the fact that this system was quite novel and not previously studied in a metacommunity framework instigated the identification of potential improvement for future studies. Because close to half of the web associate species were also found in the forest along the study transects, this suggests that the “matrix” surrounding the webs is not entirely uninhabitable. This differs from assumptions of many models, whereby the surrounding matrix between the “islands” is often barren. Future studies using this system should take this into consideration and formulate their a priori assumptions accordingly.  38  Given the importance of quantifying dispersal to identify the metacommunity model behind species distributions in any system, future research should also emphasize on measuring dispersal and quantifying population dynamics of the various associate species in a more targeted and systematic way. Moreover, since some species were actually observed associated to webs of other spider species in addition to the four host species of this study, the dispersal mechanisms might be different for each species and are probably more complex than in idealized model systems. Therefore, since the SS and ME models require more accurate dispersal (emigration/immigration) and population dynamics data, I could not attribute with certainty the species distributions to a given model. In light of the results of this study, I found that metacommunity models might not be as distinct as traditionally thought. Identifying mechanisms underlying metacommunity patterns with certainty is understandably challenging given the complexity of metacommunities in general. This thesis work should contribute to the growing number of empirical studies trying to explain such mechanisms and provide further empirical support toward the species sorting and the mass effects metacommunity models. Not unlike results in this study, a majority of published metacommunity data sets are structured by a species-sorting mechanism or by a combination of species-sorting and mass-effects processes (Cottenie 2005). The high number of studies not being able to distinguish these two metacommunity models might indicate that the framework is potentially flawed. In a paper about the terminology of metacommunity ecology, Winegardner et al. (2012) state that mass effects and patch dynamics models are actually special cases of the species-sorting model. They effectively proposed to think of metacommunities as being either neutral or species sorting, with limiting (patch dynamics), efficient (species sorting), or high dispersal  39  (mass effects). Under such a conceptual view, the mechanisms underlying community composition in this system could thus be described as being species-sorting with high dispersal, although potentially not high enough to be characteristic of a typical mass-effect model. Winegardner et al. (2012)’s framework might be a more appropriate representation of real metacommunities, as environmental filtering seems to be widely present (Burton et al. 2011; Meynard et al. 2013; Jabot et al. 2008).  In conclusion, the results in this study show a strong effect of environmental filtering in the composition of associates and low dispersal limitation and a potential combination of the species sorting and the mass effects models in explaining the species composition of associate communities of large, three-dimensional spider webs.    40  Bibliography Agnarsson, I., 2011. Habitat patch size and isolation as predictors of occupancy and number of argyrodine spider kleptoparasites in Nephila webs. Die Naturwissenschaften, 98(2), pp.163–7. Agnarsson, I., 2003. Interactive key to the world genera of cobweb spiders (Theridiidae). Agnarsson, I., 2002. Sharing a web - on the relation of sociality and kleptoparasitism in theridiid spiders (Theridiidae, Araneae). Journal of Arachnology, 30(2), pp.181–188. Alahuhta, J., Johnson, L. B., Olker, J. & Heino, J., Species sorting determines variation in the community composition of common and rare macrophytes at various spatial extents. Ecological Complexity, 20(January), pp.61–68. Almeida-Neto, M., Guimarães, P., Guimarães, P. R. & Ulrich, W., 2008. A consistent metric for nestedness analysis in ecological systems: reconciling concept and measurement. Oikos 117 (March): pp.1227–39.   Avilés, L., 1997. Causes and consequences of cooperation and permanent-sociality in spiders. The evolution of social behavior in insects and arachnids, pp.476–498. Bates, D., Maechler, M., Bolker, B. and Walker, S., 2014. Fitting linear mixed-effects   models using lme4. Journal of Statistical Software, 67(1), pp.1-48.   Beckett, S.J., Boulton, C.A. & Williams, H.T.P., 2014. FALCON: a software package for analysis of nestedness in bipartite networks. F1000Research, 3(185), pp.1–14. Borcard, D. & Legendre, P., 2002. All-scale spatial analysis of ecological data by means of principal coordinates of neighbour matrices. Ecological Modelling, 153(1-2), pp.51–68. Borcard, D., Legendre, P. & Drapeau, P., 1992. Partialling out the Spatial Component of Ecological Variation. Ecology, 73(3), pp.1045–1055. Burton, J. I., Mladenoff, D. J., Clayton, M. K. & Forrester, J. A., The roles of environmental filtering and colonization in the fine-scale spatial patterning of ground-layer plant communities in north temperate deciduous forests. Journal of Ecology, 99(3), pp.764–776. Chase, J.M. et al., 2005. Competing theories for competitive metacommunities. In M. Holyoak, M. A. Leibold, & R. D. Holt, eds. Metacommunities: spatial dynamics and ecological communities. University of Chicago Press, pp. 335–354. Cottenie, K., 2005. Integrating environmental and spatial processes in ecological community dynamics. Ecology Letters, 8(11), pp.1175–1182. Cottenie, K., Michels, E., Nuytten, N. & de Meester, L., 2003. Zooplankton metacommunity structure: regional vs . local processes in highly interconnected ponds. Ecology, 84(4), pp.991–1000. Dallas, T. & Presley, S.J., 2014. Relative importance of host environment, transmission potential and host phylogeny to the structure of parasite metacommunities. Oikos, 123(7), pp.866–874.  41  Dallas, T., 2015. metacom: Analysis of the "elements of metacommunity structure". R package version 1.4.3. Dray, S., Legendre, P. & Peres-Neto, P.R., 2006. Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecological Modelling, 196(3-4), pp.483–493. Falke, J. & Fausch, K., 2010. From metapopulations to metacommunities : linking theory with empirical observations of the spatial population dynamics of stream fishes. American Fisheries Society Symposium, pp.207–233. Fowler, H.G. & Venticinque, E.M., 1996. Interference competition and scavenging by Crematogaster ants (Hymenoptera: Formicidae) associated with the webs of the social spider Anelosimus eximius (Araneae: Theridiidae) in the Central Amazon. Journal of the Kansas Entomological Society, 69(3), pp.267–269. Guélat, J. et al., Mass effects mediate coexistence in competing shrews. Ecology, 89(7), pp.2033–2042. Guevara, J. & Avilés, L., 2011. Influence of body size and level of cooperation on the prey capture efficiency of two sympatric social spiders exhibiting an included niche pattern. Functional Ecology, 25(4), pp.859–867. Henry, T.J., 1984. Revision of the spider-commensal plant bug genus ranzovius distant (heteroptera: miridae). Proceedings of the Entomological Society of Washington, 86(February), pp.53–67. Holyoak, M., Leibold, M.A. & Holt, R.D., 2005. Metacommunities: Spatial Dynamics and Ecological Communities, Chicago: University of Chicago Press. Hubbell, S.P., 2001. The unified neutral theory of biodiversity and biogeography (MPB-32) Vol. 32., Princeton University Press. Jabot, F., Etienne, R.S., Chave J., 2008. Reconciling neutral community models and environmental filtering : theory and an empirical test. Nordic Society Oikos, 117(9), pp.1308–1320. Jimenez-Alfaro, B., Marceno, C., Guarino, R. & Chytry, M., 2015. Regional metacommunities in two coastal systems: spatial structure and drivers of plant assemblages. Journal of Biogeography, 42(3), pp.452–462. Koh, T.H. & Li, D., 2003. State-dependent prey type preferences of a kleptoparasitic spider Argyrodes flavescens (Araneae: Theridiidae). Journal of Zoology, 260(3), pp.227–233. Legendre, P. et al., 2009. Partitioning beta diversity in a subtropical broad-leaved forest of China. Ecology, 90(3), pp.663–674. Legendre, P., 1993. Spatial autocorrelation : trouble or new paradigm? Ecology, 74(6), pp.1659–1673. Legendre, P., Borcard, D. & Peres-Neto, P.R., 2005. Analyzing beta diversity: partitioning the spatial variation of community composition data. Ecological Monographs, 75(4), pp.435–450.  42  Legendre, P. & Gallagher, E.D., 2001. Ecologically meaningful transformations for ordination of species data. Oecologia, 129(2), pp.271–280. Legendre, P. & Legendre, L.F., 2012. Numerical Ecology, Vol. 24. In Developments in Environmental Modelling. Elsevier B.V. Leibold, M.A., 1998. Similarity and local co-existence of species in regional biotas. Evolutionary Ecology, 12(1), pp.95–110. Leibold, M.A. et al., 2004. The metacommunity concept: A framework for multi-scale community ecology. Ecology Letters, 7(7), pp.601–613. Leibold, M.A. & Mikkelson, G.M., 2002. Coherence, species turnover and boundary clumping: elements of meta-community structure. Oikos, 97(2), pp.237–250. Levi, H.W., 1997. The American orb weavers of the genera Mecynogea, Manogea, Kapogea and Cyrtophora (Araneae: Araneidae). Bulletin of the Museum of Comparative Zoology, 155(5), pp.215–255. Levin, S.A., 1974. Dispersion and population interactions. The American Naturalist, 108(960), pp.207–228. Levins, R., 1969. Some demographic and genetic consequences of environmental heterogeneity for biological control. Bulletin of the Entomological Society of America, 15(3), pp.237–240. Levins, R. & Culver, D., 1971. Regional Coexistence of Species and Competition between Rare Species. Proceedings of the National Academy of Sciences of the United States of America, 68(6), pp.1246–1248. Logue, J.B., Mouquet, N., Peter, H. & Hillebrand, H., 2011. Empirical approaches to metacommunities: a review and comparison with theory. Trends in ecology & evolution, 26(9), pp.482–91. MacArthur, R.H. & Wilson, E.O., 1967. The theory of island biogeography, Princeton, NJ: Princeton University Press. McGill, B.J.B., Maurer, B.A. & Weiser, M.D.M., 2006. Empirical evaluation of neutral theory. Ecology, 87(6), pp.1411–1423. Meynard, C.N., Lavergne, S., Boulangeat, I., Garraud, L., Van Es, J. & Thuiller, W., 2013. Disentangling the drivers of metacommunity structure across spatial scales. Journal of Biogeography. 40 (8): pp.1560–71. Mouquet, N. & Loreau, M., 2002. Coexistence in Metacommunities: The regional similarity hypothesis. The American Naturalist, 159(4), pp.420–426. Oksanen, J. et al., 2016. vegan: Community ecology package. R package version 2.3-4. Peres-Neto, P. R., Legendre, P., Dray, S. P. & Borcard, D., 2006. Variation partitioning of species data matrices: estimation and comparison of fractions. Ecology, 87(10), pp.2614–2625. Perkins, T.A., Riechert, S.E. & Jones, T.C., 2007. Interactions between the social spider  43  Anelosimus studiosus (Araneae, Theridiidae) and foreign spiders that frequent its nests. Journal of Arachnology, 35(1), pp.143–152. Petermann, J.S. et al., 2015. Dominant predators mediate the impact of habitat size on trophic structure in bromeliad invertebrate communities. Ecology, 96(2), pp. 428-439. Platnick, N.I., 2016. World Spider Catalog. Natural History Museum Bern. Available at: [Accessed January 1, 2016]. Platnick, N.I. & Shadab, M.U., 1978. A review of the spider genus Mysmenopsis (Araneae, Mysmenidae). American Museum Novitates, (2661), pp.1–22. Presley, S.J., Higgins, C.L. & Willig, M.R., 2010. A comprehensive framework for the evaluation of metacommunity structure. Oikos, 119(6), pp.908–917. Presley, S.J. & Willig, M.R., 2010. Bat metacommunity structure on Caribbean islands and the role of endemics. Global Ecology and Biogeography, 19(2), pp.185–199. Pruitt, J.N. & Riechert, S.E., 2011. Within-group behavioral variation promotes biased task performance and the emergence of a defensive caste in a social spider. Behavioral ecology and sociobiology, 65(5), pp.1055–1060. R Development Core Team, 2013. R: a language and environment for statistical computing. R Foundation for statistical computing, Vienna, Austria. Available at: Reeves, J.H. & Wright, D.H., 1992. On the meaning and measurement of nestedness of species assemblages. Oecologia, 92(3), pp.416–428. Romero, C. & Flórez, E., 2014. Un caso de araneofagia de Mimetus sp. (Araneae , Mimetidae), sobre Leucauge sp. (Araneae , Tetragnathidae) en el nororiente de Colombia. Revista Ibérica de Aracnología, 24(February), pp.249–250. Santos, A.J. & Brescovit, A.D., 2001. A revision of the South American spider genus Aglaoctenus Tullgren, 1905 (Araneae, Lycosidae, Sosippinae). Andrias, 15, pp.75–90. Smith, T.W. & Lundholm, J.T., 2010. Variation partitioning as a tool to distinguish between niche and neutral processes. Ecography, 33(4), pp.648–655. Srivastava, D.S. et al., 2004. Are natural microcosms useful model systems for ecology? Trends in ecology & evolution, 19(7), pp.379–84. Srivastava, D.S., 2006. Habitat structure, trophic structure and ecosystem function: interactive effects in a bromeliad-insect community. Oecologia, 149(3), pp.493–504. Staniczenko, P.P.A., Kopp, J.C. & Allesina, S., 2013. The Ghost of Nestedness in Ecological Networks. Nature Communications, 4(4), p.1391. Strona, G. et al., 2014. A fast and unbiased procedure to randomize ecological binary matrices with fixed row and column totals. Nature communications, 5(June), p.4114. Su, Y.-C. & Smith, D., 2014. Evolution of host use, group-living and foraging behaviours in kleptoparasitic spiders: molecular phylogeny of the Argyrodinae (Araneae : Theridiidae). Invertebrate Systematics, 28, pp.415–431.  44  Tilman, D., 1994. Competition and biodiversity in spatially structured habitats. Ecology, 75(1), pp.2–16. Ubick, D., Paquin, P., Cushing, P.E., Roth, V., 2005. Spiders of North America: an identification manual. Keene, New Hampshire: American Arachnological Society. Ulrich, W. & Gotelli, N.J., 2007. Null model analysis of species nestedness patterns. Ecology 88 (7): 1824–31. Ulrich, W., Almeida-Neto, M. & Gotelli, N.J., 2009. A consumer’s guide to nestedness analysis. Oikos, 118(1), pp.3–17. Vanschoenwinkel, B. et al., 2007. The role of metacommunity processes in shaping invertebrate rock pool communities along a dispersal gradient. Oikos, 116(8), pp.1255–1266. Vollrath, F., 1979. Behaviour of the kleptoparasitic spider Argyrodes elevatus (Araneae, theridiidae). Animal Behaviour, 27, pp.515–521. Wang, Y., Naumann, U., Wright, S.T. & Warton, D.I., 2012. mvabund–an R package for model‐based analysis of multivariate abundance data. Methods in Ecology and Evolution, 3(3), pp.471-474. Wheeler Jr,  a. G. & McCaffrey, J.P., 1984. Ranzovius contubernalis: seasonal history, habits, and description of fifth instar, with speculation on the origin of spider commensalism in the genus Ranzovius (Hemiptera: Miridae). Proceedings of the Entomological Society of Washington, 86(1), pp.68–81. Whitehouse, M.E.A., 1987. “ Spider Eat Spider”: The Predatory Behavior of Rhomphaea sp. from New Zealand. Journal of Arachnology, pp.355–362. Winegardner, A.K. et al., 2012. The terminology of metacommunity ecology. Trends in Ecology and Evolution, 27(5), pp.253–254. Wygodzinsky, P.W., 1966. A monograph of the Emesinae (Reduviidae, Hemiptera). Bulletin of the American Museum of Natural History, 133, pp.1–614. Yamaura, Y., Kéry, M. & Andrew Royle, J., 2016. Study of biological communities subject to imperfect detection: bias and precision of community N-mixture abundance models in small-sample situations. Ecological Research. Zheng, C., Ovaskainen, O., Roslin, T. & Tack, A. J. M., 2015. Beyond metacommunity paradigms: habitat configuration, life history, and movement shape an herbivore community on oak. Ecology, 96(12), pp.3175–3185.     45  Appendices Appendix A: Description of the four host spider species. a) Anelosimus eximius b) Anelosimus domingo c) Kapogea sexnotata and d) Aglaoctenus castaneus. Scale bars are 5mm. Tangle density is a qualitative estimate of thread density of the capture lines above the basal sheet and ranges from 1 (sparse) to 4 (dense). Host population size is the number of host spiders inhabiting the webs at the time of collection. Host body length is the mean length in millimeters of the hosts (from chelicerae to tip of abdomen). Identification from literature (Levi 1997; Santos & Brescovit 2001) and personal communications with experts.     46  Appendix B: Table of associates with photo, identification and role in the webs they inhabit. Identification from literature (Platnick & Shadab 1978; Agnarsson 2003; Platnick 2016) and personal communications with experts.   47  Appendix C: Map of the location of the webs (n = 138) collected in 2014 and 2015 (June-August). Filled symbols indicate that the host spider is social. Empty symbols indicate the host spider is solitary. 12 webs are missing from this figure due to a lack of coordinates.      


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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


Related Items