BIOTROPICA 47(5): 588–594 2015 10.1111/btp.12242
Fine-scale Beta-diversity Patterns Across Multiple Arthropod Taxa Over a Neotropical Latitudinal Gradient
Thiago Goncßalves-Souza1,9, Marcel S. Araujo 2, Eduardo P. Barbosa3, Sonia M. Lopes4, Lucas A. Kaminski5, Gustavo H. Shimizu6, Adalberto J. Santos7, and Gustavo Q. Romero8
1 Departamento de Biologia, Area de Ecologia, Universidade Federal Rural de Pernambuco (UFRPE), Rua Dom Manoel de Medeiros s/n, CEP 52171-900, Recife, PE, Brazil 2 Departamento de Zoologia e Botanica,^ Programa de Pos-Graduac ßao~ em Biologia Animal, IBILCE, Universidade Estadual Paulista, UNESP, Rua Cristov ao~ Colombo 2265, CEP 15054-000, Sao~ Jose do Rio Preto, SP, Brazil 3 Programa de Pos-Graduac ßao~ em Ecologia, Universidade Estadual de Campinas (UNICAMP), CP 6109, CEP 13083-970, Campinas, SP, Brazil 4 Museu Nacional, Setor de Blattaria, Universidade Federal do Rio de Janeiro, CEP20940-040, Rio de Janeiro, RJ, Brazil 5 Institut de Biologia Evolutiva (CSIC-Universitat Pompeu Fabra), Passeig Mar ıtim de la Barceloneta 37, 08003, Barcelona, Spain 6 Departamento de Biologia Vegetal, Programa de Pos-Graduac ßao~ em Biologia Vegetal, Universidade Estadual de Campinas (UNICAMP), CEP 13083-970, Campinas, SP, Brazil 7 Departamento de Zoologia, Instituto de Ciencias^ Biologicas, Universidade Federal de Minas Gerais (UFMG),Av. Antonio^ Carlos 6627, CEP 31270-901, Belo Horizonte, MG, Brazil 8 Departamento de Biologia Animal, Instituto de Biologia, Universidade Estadual de Campinas (UNICAMP), CP 6109, CEP 13083-970, Campinas, SP, Brazil
ABSTRACT Documenting how diversity patterns vary at fine- and broad scales may help answer many questions in theoretical and applied ecology. However, studies tend to compare diversity patterns at the same scale and within the same taxonomic group, which limits the applicabil- ity and generality of the results. Here, we have investigated whether vegetation-dwelling arthropods from different trophic ranks and with distinct life histories (i.e., ants, caterpillars, cockroaches, and spiders) have different beta-diversity patterns at multiple scales. Specifi- cally, we compared their beta diversity across architecturally distinct plant species (fine-scale process) and a latitudinal gradient of sites (broad-scale process) along 2040 km of coastal restinga vegetation in the Neotropics. Over 50 percent of the compositional changes (b-diversity) in ants, caterpillars, and spiders and 41 percent of those in cockroaches were explained by plant identity within each site. Even groups that do not feed on plant tissues, such as omnivores and predators, were strongly affected by plant identity. Fine-scale vari- ation was more important than large-scale processes for all studied groups. Performing a cross-scale comparison of diversity patterns of groups with distinct life histories helps elucidate how processes that act at regional scales, such as dispersal, interact with local processes to assemble arthropod communities.
Abstract in Portuguese is available with online material.
Key words: alpha, beta, and gamma diversity; dispersal; diversity partitioning; life history; plant identity.
BIOTIC INTERACTIONS, ENVIRONMENTAL FILTERS (E.G., HABITAT STRUC- and physiological requirements (Crist et al. 2006). However, most TURE), SPATIAL SCALE, AND LANDSCAPE CONFIGURATION ARE THE studies of arthropods have been performed at similar scales and MAIN DRIVERS OF ARTHROPOD SPECIES COMPOSITION (Basset 1996, on organisms with similar requirements, limiting the applicability Lewinsohn et al. 2005, Schaffers et al. 2008). The relative impor- of the results to other groups (but see Shurin et al. 2009). By tance of these factors for species abundance and composition considering species with different requirements (e.g., food, habitat, appears to be scale and taxon dependent (Steffan-Dewenter et al. and climatic preferences), one can discern not only general pat- 2002, Soininen et al. 2007a). Consequently, differences in the terns shared by ecologically distinct arthropod groups but also characteristics and spatial distribution of a habitat should be per- the idiosyncratic responses of those groups to varying environ- ceived differently depending on an animal’s behavioral, ecological, mental conditions. In addition to environmental requirements, the way in which Received 15 December 2014; revision accepted 14 April 2015. organisms disperse (actively or passively) can reliably predict 9Corresponding author; e-mail: [email protected] how broad- and fine-scale processes determine species turnover 588 ª 2015 The Association for Tropical Biology and Conservation Beta-diversity of Multiple Arthropod Taxa 589
(i.e., compositional changes in communities between two sampling organisms (e.g., ants, cockroaches, spiders), while not herbivorous, units, also known as b-diversity; Tuomisto 2010, Anderson et al. nevertheless depend on plant architecture during various stages 2011). We considered two broad dispersal modes, i.e., active dis- of their life cycle. For instance, plants are an important resource persers that fly and passive dispersers that cannot fly but may utilized by arthropods for web building, foraging, mating, and use wind to disperse (e.g., spider ballooning). At broad scales, dis- oviposition as well as obtaining protection against desiccation and persal ability determines a species’ spatial distribution and its natural enemies (Romero & Vasconcellos-Neto 2005 and refer- probability of finding suitable conditions to feed and reproduce ences therein). Indeed, plant species composition and architecture (Kisdi 2002). Generally, active dispersal favors the spread of an affect the composition of omnivorous and predatory arthropod organism and the likelihood of its finding suitable habitat condi- communities (Halaj et al. 2000, Haddad et al. 2001, Johnson et al. tions, whereas passive dispersal is more stochastic and depends, 2006, Goncßalves-Souza et al. 2010). In addition, plant architecture for example, on weather conditions (Soininen et al. 2007a,b). As and composition are thought to have stronger effects on the a result, at broad spatial scales, species turnover of passive dis- community structure of spiders relative to other predatory and persers is expected to be higher than the turnover of active dis- omnivorous arthropods (e.g., Schaffers et al. 2008). However, persers, suggesting that the dispersal mode affects b-diversity detritivore communities may be less affected by plant species patterns (Tuomisto et al. 2003, Soininen et al. 2007a). Conversely, composition: Siemann et al. (1998) showed that detritivores are dispersal may be less important at finer scales; thus, local ecologi- the only plant-dwelling trophic group not affected by plant spe- cal factors such as competition (Graham et al. 2009) and habitat cies richness. As a result, we predict that changes in plant identity suitability (Brehm et al. 2003, Schaffers et al. 2008, Goncßalves- more strongly affect the composition of herbivore communities Souza et al. 2010) might account for compositional changes at than non-herbivore communities. Additionally, these effects these scales. Testing how species turnover varies in response to should be more intense for predators than for omnivores and broad-scale processes (e.g., among sites in a Neotropical latitudi- detritivores. nal gradient) and fine-scale processes (e.g., in different microhabi- In this study, we assessed whether the a- and b-diversity of tats) could elucidate a variety of mechanisms that drive species multiple taxa of plant-dwelling arthropods (ants, cockroaches, distribution. caterpillars, and spiders) varied at different spatial scales and with Few studies have compared how local ecological factors, different plant species. In addition, we viewed ants (via the flight such as microhabitat variations, affect taxa from different guilds of reproductive females), caterpillars (via adult flight), and cock- (Crist et al. 2006). The current consensus suggests that plant spe- roaches as active dispersers (Peeters & Ito 2001, Gillott 2005, cies composition has a stronger influence on herbivore b-diversity Bell et al. 2007) and spiders as passive dispersers (Bell et al. 2005, (Novotny et al. 2004, Prado & Lewinsohn 2004, M€uller et al. Foelix 2011). Specifically, we investigated to what extent plant 2011, Kitching et al. 2013) than on the b-diversity of predators identity and among-location variation cause differences in or omnivores. Most studies have revealed that plant defenses are b-diversity patterns among ants, caterpillars, cockroaches, and spi- the main factors driving this host specificity in herbivores (re- ders. We made two predictions (see Fig. 1): (1) among different viewed in Coley & Barone 1996). However, many plant-dwelling plant species within each locality, the b-diversity pattern will be
FIGURE 1. Schematic illustration of the study design with the two scales used to compare b-diversity. The fine scale considers arthropod species turnover among b different plant species ( 1). The various tree-like symbols illustrate plants with distinct architectures. The broad scale considers species turnover among different a a sites (gray rectangles). 1 represents the value within each individual plant, and 2 represents the average diversity within each plant species. We partitioned total diversity (c) as implemented by Jost (2007) (formula at the bottom of the figure). 590 Goncßalves-Souza et al.
b < b b fi fi non-herbivores herbivores since -diversity is related to the speci- we found ve individuals of any plant species in the rst four ficity of food consumption in each arthropod group (Novotny & plots, we did not sample arthropods on this species in the follow- Weiblen 2005, Schaffers et al. 2008). Additionally, host plant char- ing plots because the maximum number of individual plants per acteristics or identity (Omena & Romero 2010) will cause stron- architecture type per location was set to 20. We sampled the ger variations in the b-diversity of predators compared to same plots in both years but with a randomized sampling b b detritivores. (2) Among different sites, the -diversity will be ac- sequence. tive < bpassive because active dispersers are expected to be less We collected arthropods (Araneae, Blattaria, Hymenoptera: restricted by geographical distance than passive dispersers (spi- Formicidae, and Lepidoptera: exophagous caterpillars) on the ders). However, the specificity of lepidopterans to their host branches of each plant using the following protocol: (1) we plants may be more important than dispersal, leading to higher packed four to ten branches (depending on branch size) in 100 L b-diversity among sites (Komonen et al. 2004). transparent plastic bags and cut off the branches; (2) we carefully shook the bag 20 times to release arthropods from the branches; METHODS (3) we removed each branch to check for arthropods in a white tray; (4) we collected every arthropod visible to the naked eye STUDY AREA.—We sampled arthropods in 12 localities (from lati- found on the branches and inside the bags and preserved them tudes 12S to 28S) embedded in the restinga vegetation along in 75 percent ethanol. This method was repeated for each indi- 2040 km of the Brazilian coast. Climate data for each locality are vidual plant. We implemented another protocol for bromeliads, shown in Appendix S1. The restinga is a shrubby vegetation for- i.e., collecting arthropods (visible to the naked eye) on the entire mation distributed along sandy plains formed by late Quaternary plant surface, except for aquatic taxa living within reservoirs. In marine deposition in eastern Brazil (Scarano 2002). We choose a addition, we counted the number of leaves and weighed three specific type of restinga, called ‘open restinga’, which consists of leaves (the smallest leaf, an intermediate-sized leaf, and the largest patchy vegetation surrounded by open areas covered with sand leaf). We then multiplied the number of leaves by the average of or herbaceous vegetation (Scarano 2002). The main plant families the three weighed leaves to estimate total leaf biomass. in open restingas are Arecaceae, Bromeliaceae, Malpighiaceae, We categorized ants, caterpillars, cockroaches, and spiders as Myrtaceae, Rubiaceae, and Sapindaceae (Assis et al. 2004). omnivores, herbivores, generalist detritivores, and predators, respectively (Gillott 2005, Bell et al. 2007, Foelix 2011), excluding PLANT SPECIES.—We selected five plant species at each site based predatory and fungivorous ants from the analysis. We examined on leaf characteristics and crown size. As we cannot separate the specimens using a stereoscopic microscope and sorted them into effects of plant architecture from other confounding effects, such morphospecies. We deposited ants in the Laboratorio de Inter- as plant chemistry, on species a and b diversity, we use the term acßoes~ Animais-Plantas, Universidade Estadual de Campinas (T.M. “plant identity” to represent the architectural characteristics rele- Lewinsohn), caterpillars in the Laboratorio de Borboletas, Univer- vant to prediction 1. Specifically, at each locality, we selected a sidade Estadual de Campinas (A.V.L. Freitas), cockroaches in the bromeliad (Bromeliaceae), a palm (Arecaceae), and three different Blattaria collection of Museu Nacional do Rio de Janeiro (S.M. dicot plants with small (leaf length: 3.01 1.43 mean SD; leaf Lopes), and spiders in the arachnid collection of Colecßoes~ width: 1.51 0.72), medium (leaf length: 6.63 2.08; leaf Taxonomicas^ da Universidade Federal de Minas Gerais (A.J. width: 2.83 0.90), and large leaves (leaf length: 11.45 2.54; Santos). leaf width: 6.48 1.75). For the four localities without palms, we selected another common dicot plant distinct from bromeliads STATISTICAL ANALYSES.—We used a multiplicative Diversity Parti- and from the three other dicots. We did not collect flowering tioning method to decompose c diversity into independent a and plants or plants bearing extrafloral nectaries (EFNs) to avoid col- b components from numbers equivalents (sensu Jost 2007). Alpha lecting species that are specialized to exploit these resources. For and beta components of the numbers equivalents were calculated instance, spiders of the family Thomisidae are highly specialized by: flower-dwelling organisms (Morse 2007), and ants are the most frequent visitors of EFN-bearing plants (Oliveira et al. 1999). DðHcÞ¼DðHaÞDðHbÞ;
ARTHROPOD SURVEY.—We collected arthropods on the plant spe- where D is the numbers equivalent of each diversity measure cies in each locality between September and November 2009 and (Shannon diversity in our case: q = 1; see below). Using the for- from June through August 2010. We sampled 20 individuals of mula derived by Jost (2007), comparisons of a and b compo- each plant species (architecture type) in 20 plots (30 m 9 30 m) nents among taxonomic groups and regions with distinct within each locality. The plots were separated from each other by numbers of species are equivalent (Chao et al. 2012). This prop- at least 50 m. The criterion for choosing plots was the presence erty is based on Hill numbers, and Jost (2007) calculates the true of at least three of the five plant architectures. We randomized diversity using the parameter q, which gives weights to common the order of plot sampling. Within each plot, we sampled up to or rare species. In particular, we set q = 1 to avoid favoring five individual plants of each architecture type to increase the either rare or common species (Jost 2007, Chao et al. 2012) and spatial distribution of plants sampled within the same locality. If because this value eliminates the effect of sample size on the Beta-diversity of Multiple Arthropod Taxa 591
results. The use of numbers equivalents is necessary because the traditional diversity indices are superfluous and because the parti- tioning of diversity based on these indices does not yield inde- pendent components (Jost 2007, Chao et al. 2012). Here, the term b-diversity refers to compositional changes among distinct sampling units (see Tuomisto 2010). Each individual plant represents the lowest hierarchical level, followed by among-plant species and among-site comparisons. Thus, we calculated a-diversity at two levels: a-diversity within a a individual plants ( 1) and the average -diversity within each plant a b species ( 2). The components of -diversity can be divided into different levels that reflect different spatial scales (Crist et al. 2003, Fig. 1). Because individual plants were nested within five different plant species and plant species were nested within differ- ent sites, we partitioned b-diversity into two components: b-di- versity among plant species (b ) and among sites (b ) (Fig. 1). 1 2 FIGURE 2. c diversity of each arthropod group was partitioned into a We considered these two b-diversity components to represent 1 (within individual plants), a (within plant species), b (among plant species), fine-scale and broad-scale patterns in species compositional dif- 2 1 and b (among sites) components. The values within bars represent the per- ferences. We used an individual-based randomization scheme 2 centage of variation explained by each partitioned value in relation to c diver- (which generated 1000 random matrices with fixed rows and col- sity. Error bars were generated using a bootstrap procedure with 999 umn totals) to compare the observed and expected values of the iterations. The observed values of each component in all analyses were signifi- partitioned components. Additionally, we used a bootstrap proce- cantly different from the value expected by chance (P < 0.05). dure with 999 iterations to estimate uncertainty around each par- titioned component. We pooled species compositional data for all studied arthro- active dispersers. The partitioning of Shannon diversity was sig- pods from the 2 years because they did not vary (results not nificantly lower than the null expectation for all levels and for all shown). We repeated the diversity partitioning method for each taxonomic groups (P < 0.05). arthropod group. All statistical analyses were performed in R 2.15 (R Development Core Team 2012) with the vegan package DISCUSSION (Oksanen et al. 2009). Our results reveal that fine-scale changes in plant architecture RESULTS may be better predictors of arthropod b-diversity than broad- scale changes (turnover among sites). In general, b-diversity is We found 856 ants from 48 species, 766 caterpillars from 161 negatively related to dispersal ability because high dispersal species, 172 cockroaches from 32 species, and 1122 spiders from homogenizes the spatial distribution of species (Mouquet & Lor- 172 species. On average, we found 5.3 arthropod species per eau 2003, Soininen et al. 2007b). However, spiders (which dis- plant. Caterpillars and spiders were the groups with the most spe- perse passively) were the group with the lowest b-diversity b cies per individual plant, followed by cockroaches and ants. among sites ( 2), followed by ants, caterpillars, and cockroaches. a a The value of -diversity within an individual plant ( 1) and Shurin et al. (2009) found lower b-diversity in organisms that dis- a the average value within each plant species ( 2) together repre- perse passively (e.g., plankton) and suggested that differences in sented between 29 and 36 percent of the total diversity (Fig. 2). life history and dispersal strategies may generate similar levels of The turnover among plant species explained a greater proportion dispersal limitation and, consequently, similar b-diversity among c (more than 50%) of the -diversity for caterpillars, spiders, and sites (as in the case of b2 for ants, caterpillars, and spiders). ants and 41 percent of the c-diversity for cockroaches (Fig. 2). Although adult lepidopterans are active dispersers, their mobility
That is, b1(cockroaches) < b1(ants)